chore: release v0.4.4 on master

A Scheduled Task stores an absolute path. Checking that a task named
"OSTP TUN Helper" exists said nothing about whether its <Command> still points
at the helper we are about to run, and the paths do drift: a dev build
registers target\debug\ostp-tun-helper.exe, an installer registers Program
Files, and moving or reinstalling the app leaves the old path behind.

That failed silently in the worst way. schtasks /Run reports success for
merely ACCEPTING the launch request — a task whose exe no longer exists fails
afterwards, out of band, with nothing returned to us. So launch_as_admin
returned Ok, and the caller then sat in its 60-second connect loop before
reporting "Timeout connecting to helper." On every connect, permanently, with
no way out except deleting the task by hand.

The check now reads the registered <Command> back and compares it to the exe,
re-registering through the existing /F overwrite when they differ: one consent
prompt, once, instead of a permanent silent breakage.

The path is read via /Query /XML rather than /FO LIST /V because the list
format's field labels are localized — "Task To Run" is "Задача для запуска" on
a Russian Windows — while XML tag names are not. schtasks emits UTF-16LE with
a BOM there, which is decoded explicitly, with UTF-8 tolerated as a fallback.
Both paths are canonicalized before comparison so casing, `..` and 8.3 short
names do not read as a mismatch; a path that cannot be canonicalized no longer
exists, which is itself grounds to re-register.

.github/workflows/release.yml

@@ -1,212 +1,726 @@
-name: Universal CI/CD Release Matrix
+name: CI/CD
+
+
+# `run-name` is evaluated at workflow-start, BEFORE any job runs - it cannot
+# see resolve-channel's computed tag_name (e.g. "0.4.3-alpha"), only the
+# `github.*` context. The old "release version ${{ github.ref_name }}" showed
+# the bare branch name ("alpha"/"beta") for every run, which reads
+# exactly like a literal release tag and caused real confusion - the actual
+# release tag has been correct (versioned) all along; only this label lied
+# about it. Spell out "channel" so nobody mistakes one for the other again.
+# NOTE: this value MUST be quoted. The GHA string literal below contains
+# "Release build: {0}" - an unquoted YAML plain scalar treats ": " (colon
+# then space) as starting a nested mapping, which is exactly what broke every
+# single push since this line was introduced: GitHub rejected the whole
+# workflow file at parse time (before any job runs), silently burning an
+# Actions-minutes-billed run per push for nothing.
+run-name: "${{ startsWith(github.ref, 'refs/tags/') && (contains(github.ref_name, 'beta') && format('CI/CD: beta version {0}', github.ref_name) || contains(github.ref_name, 'alpha') && format('CI/CD: alpha version {0}', github.ref_name) || format('CI/CD: release version {0}', github.ref_name)) || format('CI/CD: {0} channel build', github.ref_name) }}"
 
 on:
   push:
     tags:
       - "v*"
   workflow_dispatch:
+    inputs:
+      channel:
+        description: >-
+          Manually build+release just this rolling channel. Stable releases
+          are NEVER picked here on purpose - cut those only via a real
+          "vX.Y.Z" tag push, so a manual dispatch can't accidentally publish
+          a "stable" release.
+        type: choice
+        required: true
+        default: alpha
+        options:
+          - alpha
+          - beta
 
 permissions:
   contents: write
 
+# -- Global defaults ---------------------------------------------------------
+env:
+  CARGO_TERM_COLOR: always
+  CARGO_INCREMENTAL: 0
+  RUST_BACKTRACE: short
+
 jobs:
+  # Computes ONE channel + release tag for this whole run, so every build
+  # job (native matrix + all 3 GUI platforms + Android) uploads to the exact
+  # same release under the exact same tag, instead of repeating this logic
+  # (and risking it drifting out of sync) in five separate places.
+  #
+  # Tag shape:
+  #   - real "vX.Y.Z" / "vX.Y.Z-beta.N" tag push -> tag used as-is (stable promotion)
+  #   - push to `alpha`                        -> "{version}-alpha" (rolling, same tag every push)
+  #   - push to `beta`                            -> "{version}-beta"    (rolling, same tag every push)
+  #   - workflow_dispatch                         -> forced by the `channel` input (alpha|beta only)
+  resolve-channel:
+    name: Resolve release channel
+    runs-on: ubuntu-latest
+    outputs:
+      channel: ${{ steps.resolve.outputs.channel }}
+      tag_name: ${{ steps.resolve.outputs.tag_name }}
+      prerelease: ${{ steps.resolve.outputs.prerelease }}
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Resolve channel, version, and release tag
+        id: resolve
+        shell: bash
+        run: |
+          set -euo pipefail
+          BASE_VERSION=$(grep -m1 '^version' Cargo.toml | sed -E 's/version *= *"([^"]+)"/\1/')
+
+          if [[ "${{ github.ref }}" == refs/tags/v* ]]; then
+            # A pushed tag is authoritative — use it AS-IS (never recompute it
+            # from Cargo.toml, or the release would upload to a different tag than
+            # the one that triggered this run). The channel, and thus prerelease,
+            # is decided by the tag's suffix: v0.4.7-beta / v0.4.7-alpha are
+            # prereleases; a bare vX.Y.Z is the only thing that becomes stable.
+            TAG="${{ github.ref_name }}"
+            case "$TAG" in
+              *-alpha*) CHANNEL="alpha" ;;
+              *-beta*)  CHANNEL="beta"  ;;
+              *)        CHANNEL="stable" ;;
+            esac
+          else
+            # No tag (workflow_dispatch, or a legacy branch push): pick the
+            # channel, then synthesize the rolling tag from Cargo.toml's version.
+            if [ "${{ github.event_name }}" = "workflow_dispatch" ]; then
+              CHANNEL="${{ github.event.inputs.channel }}"
+            elif [ "${{ github.ref_name }}" = "beta" ]; then
+              CHANNEL="beta"
+            else
+              CHANNEL="alpha"
+            fi
+            TAG="v${BASE_VERSION}-${CHANNEL}"
+          fi
+
+          echo "Resolved channel=$CHANNEL tag=$TAG (base version $BASE_VERSION)"
+          echo "channel=$CHANNEL" >> "$GITHUB_OUTPUT"
+          echo "tag_name=$TAG" >> "$GITHUB_OUTPUT"
+          echo "prerelease=$([ "$CHANNEL" = "stable" ] && echo false || echo true)" >> "$GITHUB_OUTPUT"
+
+  check-and-test:
+    name: Check & Test
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+
+      - name: Setup Rust toolchain
+        uses: dtolnay/rust-toolchain@stable
+        with:
+          toolchain: stable
+
+      - name: Restore Cargo cache
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cargo/registry/index/
+            ~/.cargo/registry/cache/
+            ~/.cargo/git/db/
+            target/
+          key: cargo-check-${{ hashFiles('**/Cargo.lock') }}
+          restore-keys: cargo-check-
+
+      - name: Install musl-tools
+        run: sudo apt-get update && sudo apt-get install -y musl-tools
+
+      - name: Create dummy dist for rust-embed
+        run: mkdir -p ostp-control/dist && touch ostp-control/dist/index.html
+
+      - name: cargo check
+        run: cargo check --workspace
+
+      - name: cargo test
+        run: cargo test --workspace --lib
+
   publish-release-matrix:
     name: Release for ${{ matrix.target }}
+    needs: [check-and-test, resolve-channel]
     runs-on: ${{ matrix.os }}
     strategy:
       fail-fast: false
       matrix:
         include:
-          # ==========================================
-          # 🏁 WINDOWS ECOSYSTEM
-          # ==========================================
+          # -- Windows ------------------------------------------------------
           - os: windows-latest
             target: x86_64-pc-windows-msvc
             artifact_name: ostp.exe
             release_name: ostp-windows-amd64.zip
-            tun2socks_arch: windows-amd64
             wintun_arch: amd64
+
           - os: windows-latest
             target: i686-pc-windows-msvc
             artifact_name: ostp.exe
             release_name: ostp-windows-386.zip
-            tun2socks_arch: windows-386
             wintun_arch: x86
+
           - os: windows-latest
             target: aarch64-pc-windows-msvc
             artifact_name: ostp.exe
             release_name: ostp-windows-arm64.zip
-            tun2socks_arch: windows-arm64
             wintun_arch: arm64
 
-          # ==========================================
-          # 🍏 APPLE DARWIN (macOS)
-          # ==========================================
+          # -- macOS ---------------------------------------------------------
           - os: macos-latest
             target: x86_64-apple-darwin
             artifact_name: ostp
             release_name: ostp-darwin-amd64.tar.gz
-            tun2socks_arch: darwin-amd64
+
           - os: macos-latest
             target: aarch64-apple-darwin
             artifact_name: ostp
             release_name: ostp-darwin-arm64.tar.gz
-            tun2socks_arch: darwin-arm64
 
-          # ==========================================
-          # 🐧 LINUX & FreeBSD STANDARD
-          # ==========================================
+          # -- Linux native --------------------------------------------------
           - os: ubuntu-latest
             target: x86_64-unknown-linux-musl
             artifact_name: ostp
             release_name: ostp-linux-amd64.tar.gz
-            tun2socks_arch: linux-amd64
+
           - os: ubuntu-latest
             target: i686-unknown-linux-musl
             artifact_name: ostp
             release_name: ostp-linux-386.tar.gz
-            tun2socks_arch: linux-386
+            use_cross: true
+
+          # -- Linux cross ---------------------------------------------------
           - os: ubuntu-latest
             target: aarch64-unknown-linux-musl
             artifact_name: ostp
             release_name: ostp-linux-arm64.tar.gz
-            tun2socks_arch: linux-arm64
             use_cross: true
+
           - os: ubuntu-latest
             target: armv7-unknown-linux-musleabihf
             artifact_name: ostp
             release_name: ostp-linux-armv7.tar.gz
-            tun2socks_arch: linux-armv7
             use_cross: true
+
           - os: ubuntu-latest
             target: x86_64-unknown-freebsd
             artifact_name: ostp
             release_name: ostp-freebsd-amd64.tar.gz
-            tun2socks_arch: freebsd-amd64
             use_cross: true
 
-          # ==========================================
-          # 🛰️ ROUTER & SPECIAL ARCHITECTURES (Cross)
-          # ==========================================
           - os: ubuntu-latest
             target: mipsel-unknown-linux-musl
             artifact_name: ostp
             release_name: ostp-linux-mipsle.tar.gz
-            tun2socks_arch: linux-mipsle
             use_cross: true
             toolchain: nightly
+
           - os: ubuntu-latest
             target: riscv64gc-unknown-linux-gnu
             artifact_name: ostp
             release_name: ostp-linux-riscv64.tar.gz
-            tun2socks_arch: linux-riscv64
             use_cross: true
 
-          # ==========================================
-          # 🤖 MOBILE & EMBEDDED SUITE (Cross)
-          # ==========================================
-          - os: ubuntu-latest
-            target: aarch64-linux-android
-            artifact_name: ostp
-            release_name: ostp-android-arm64.tar.gz
-            tun2socks_arch: linux-arm64
-            use_cross: true
+
 
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
 
-      - name: Initialize Rust ecosystem (Native Host)
-        if: ${{ !matrix.use_cross }}
+      # -- Frontend Build -----------------------------------------------------
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 20
+      - name: Build Web Panel (skip if no source; use committed dist/)
+        shell: bash
+        run: |
+          mkdir -p ostp-control/dist
+          cd ostp-control
+          if [ -f package.json ]; then
+            npm install && npm run build
+          else
+            echo "ostp-control has no package.json - using committed dist/"
+            [ -f dist/index.html ] || echo '<!doctype html><title>OSTP</title>' > dist/index.html
+          fi
+
+      # -- Rust toolchain -----------------------------------------------------
+      - name: Setup Rust toolchain
         uses: dtolnay/rust-toolchain@stable
         with:
           toolchain: ${{ matrix.toolchain || 'stable' }}
-          targets: ${{ matrix.target }}
+          targets: ${{ !matrix.use_cross && matrix.target || '' }}
 
-      - name: Initialize Rust ecosystem (Cross Container Host)
-        if: ${{ matrix.use_cross }}
-        uses: dtolnay/rust-toolchain@stable
+      # -- Cargo cache (shared per target) -----------------------------------
+      - name: Restore Cargo cache
+        uses: actions/cache@v4
         with:
-          toolchain: ${{ matrix.toolchain || 'stable' }}
+          path: |
+            ~/.cargo/registry/index/
+            ~/.cargo/registry/cache/
+            ~/.cargo/git/db/
+            target/
+          key: cargo-${{ matrix.target }}-${{ hashFiles('**/Cargo.lock') }}
+          restore-keys: |
+            cargo-${{ matrix.target }}-
 
-      - name: Activate rust compilation caching
-        if: ${{ !matrix.use_cross }}
-        uses: swatinem/rust-cache@v2
-
-      - name: Setup local MUSL linker dependencies
+      # -- MUSL tools for native Linux musl builds ----------------------------
+      - name: Install musl-tools
         if: ${{ matrix.os == 'ubuntu-latest' && !matrix.use_cross }}
         run: sudo apt-get update && sudo apt-get install -y musl-tools
 
-      - name: Execute Standard Native Compilation (Windows)
-        if: ${{ !matrix.use_cross && matrix.os == 'windows-latest' }}
+      # -- Native build -------------------------------------------------------
+      - name: Build (native)
+        if: ${{ !matrix.use_cross }}
+        shell: bash
         run: cargo build --release --target ${{ matrix.target }} --bin ostp
 
-      - name: Execute Standard Native Compilation (Unix)
-        if: ${{ !matrix.use_cross && matrix.os != 'windows-latest' }}
-        shell: bash
-        run: |
-          cargo build --release --target ${{ matrix.target }} --bin ostp
-
-      - name: Execute Specialized Cross-Compilation
+      # -- Cross build --------------------------------------------------------
+      - name: Restore cross binary cache
         if: ${{ matrix.use_cross }}
-        run: |
-          cargo install cross --git https://github.com/cross-rs/cross.git
-          cross build --release --target ${{ matrix.target }} --bin ostp
+        id: cross-cache
+        uses: actions/cache@v4
+        with:
+          path: ~/.cargo/bin/cross
+          key: cross-bin-${{ runner.os }}-v1
 
-      - name: Inject Win32 Driver Dependencies (Windows)
-        if: ${{ matrix.os == 'windows-latest' && matrix.tun2socks_arch }}
-        shell: pwsh
-        run: |
-          cd target/${{ matrix.target }}/release
-          $ProgressPreference = 'SilentlyContinue'
-          # 1. Acquire tun2socks
-          Invoke-WebRequest -Uri "https://github.com/xjasonlyu/tun2socks/releases/download/v2.6.0/tun2socks-${{ matrix.tun2socks_arch }}.zip" -OutFile "tun2socks.zip"
-          Expand-Archive -Path "tun2socks.zip" -DestinationPath "tun_temp" -Force
-          Get-ChildItem -Path "tun_temp" -Filter "*.exe" -Recurse | Copy-Item -Destination "." -Force
-          # 2. Acquire wintun
-          Invoke-WebRequest -Uri "https://www.wintun.net/builds/wintun-0.14.1.zip" -OutFile "wintun.zip"
-          Expand-Archive -Path "wintun.zip" -DestinationPath "wintun_temp" -Force
-          Get-ChildItem -Path "wintun_temp" -Filter "wintun.dll" -Recurse | Where-Object { $_.FullName -match 'bin[\\/]${{ matrix.wintun_arch }}[\\/]' } | Copy-Item -Destination "." -Force
-          # Cleanup
-          Remove-Item "tun2socks.zip", "tun_temp", "wintun.zip", "wintun_temp" -Recurse -Force
+      - name: Install cross (if not cached)
+        if: ${{ matrix.use_cross && steps.cross-cache.outputs.cache-hit != 'true' }}
+        # cross-rs's own source (not ours, not a dependency of ours) uses a
+        # macro-at-end-of-block pattern that trips rustc's
+        # semicolon_in_expressions_from_macros lint on current toolchains -
+        # harmless in cross's actual behavior, but `cargo install` compiles
+        # the installed package as the "local" crate, so dependency lint
+        # capping doesn't shield it. --cap-lints=warn is the standard escape
+        # hatch for building a third-party tool against a newer compiler than
+        # its own lint config assumed; it doesn't touch our own build.
+        run: RUSTFLAGS="--cap-lints=warn" cargo install cross --git https://github.com/cross-rs/cross.git --locked
 
-      - name: Inject Unix Driver Dependencies (Unix)
-        if: ${{ matrix.os != 'windows-latest' && matrix.tun2socks_arch }}
-        shell: bash
-        run: |
-          cd target/${{ matrix.target }}/release
-          # All platforms in tun2socks v2.6.0 use .zip packaging
-          URL="https://github.com/xjasonlyu/tun2socks/releases/download/v2.6.0/tun2socks-${{ matrix.tun2socks_arch }}.zip"
-          curl -L "$URL" -o "tun2socks.zip"
-          unzip -o "tun2socks.zip"
-          find . -maxdepth 2 -name "tun2socks*" ! -name "*.zip" -type f -exec mv {} ./tun2socks \;
-          rm -f "tun2socks.zip"
-          chmod +x tun2socks || true
+      - name: Build (cross)
+        if: ${{ matrix.use_cross }}
+        run: cross build --release --target ${{ matrix.target }} --bin ostp
 
-      - name: Package release artifact (Windows)
+      # -- Driver dependencies ------------------------------------------------
+      - name: Download wintun (Windows)
         if: ${{ matrix.os == 'windows-latest' }}
         shell: pwsh
         run: |
-          cd target/${{ matrix.target }}/release
-          $files = @("${{ matrix.artifact_name }}")
-          if (Test-Path "tun2socks.exe") { $files += "tun2socks.exe" }
-          if (Test-Path "wintun.dll") { $files += "wintun.dll" }
-          Compress-Archive -Path $files -DestinationPath ../../../${{ matrix.release_name }}
+          $ProgressPreference = 'SilentlyContinue'
+          $dir = "target/${{ matrix.target }}/release"
+          Invoke-WebRequest -Uri "https://www.wintun.net/builds/wintun-0.14.1.zip" -OutFile "$dir/wt.zip"
+          Expand-Archive "$dir/wt.zip" -DestinationPath "$dir/wt_tmp" -Force
+          Get-ChildItem "$dir/wt_tmp" -Filter "wintun.dll" -Recurse | Where-Object { $_.FullName -match 'bin[\\/]${{ matrix.wintun_arch }}[\\/]' } | Copy-Item -Destination "$dir/"
+          Remove-Item "$dir/wt.zip","$dir/wt_tmp" -Recurse -Force
 
-      - name: Package release artifact (Unix Systems)
+      # -- Package ------------------------------------------------------------
+      - name: Package (Windows)
+        if: ${{ matrix.os == 'windows-latest' }}
+        shell: pwsh
+        run: |
+          $dir = "target/${{ matrix.target }}/release"
+          $files = @("ostp.exe")
+          if (Test-Path "$dir/wintun.dll")    { $files += "wintun.dll" }
+          Push-Location $dir
+          Compress-Archive -Path $files -DestinationPath "../../../${{ matrix.release_name }}" -Force
+          Pop-Location
+
+      - name: Package (Unix)
         if: ${{ matrix.os != 'windows-latest' }}
         run: |
-          cd target/${{ matrix.target }}/release
+          dir="target/${{ matrix.target }}/release"
           FILES="${{ matrix.artifact_name }}"
-          if [ -f "tun2socks" ]; then
-            FILES="$FILES tun2socks"
-          fi
-          tar -czf ../../../${{ matrix.release_name }} $FILES
+          tar -czf "${{ matrix.release_name }}" -C "$dir" $FILES
 
-      - name: Inject artifact to Global GitHub Release Assets
-        if: ${{ startsWith(github.ref, 'refs/tags/') }}
+      # -- Upload -------------------------------------------------------------
+      - name: Upload to GitHub Release
         uses: softprops/action-gh-release@v2
         with:
+          # Computed once in resolve-channel so every platform/job in this run
+          # lands on the exact same tag: "{version}-alpha" / "{version}-beta"
+          # for rolling channel pushes, or the pushed "vX.Y.Z" tag as-is for a
+          # real stable release.
+          tag_name: ${{ needs.resolve-channel.outputs.tag_name }}
+          prerelease: ${{ needs.resolve-channel.outputs.prerelease }}
           files: ${{ matrix.release_name }}
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+
+  build-windows-gui:
+    name: Build Windows GUI (Tauri) - ${{ matrix.arch }}
+    needs: [check-and-test, resolve-channel]
+    runs-on: windows-latest
+    strategy:
+      matrix:
+        include:
+          - arch: amd64
+            target: x86_64-pc-windows-msvc
+          - arch: arm64
+            target: aarch64-pc-windows-msvc
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 20
+
+      - name: Setup Rust toolchain
+        uses: dtolnay/rust-toolchain@stable
+        with:
+          targets: ${{ matrix.target }}
+
+      - name: Install Tauri CLI
+        run: npm install -g @tauri-apps/cli
+
+      - name: Cache cargo
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cargo/registry/index/
+            ~/.cargo/registry/cache/
+            ~/.cargo/git/db/
+            target/
+            ostp-gui/src-tauri/target/
+          key: cargo-windows-gui-${{ matrix.target }}-${{ hashFiles('**/Cargo.lock') }}
+          # Without a prefix fallback this cache NEVER restored on a release:
+          # cutting a release rewrites every Cargo.lock (version bump), which
+          # changes hashFiles(), which misses the exact key — so each release
+          # rebuilt every dependency from scratch. That is why the GUI jobs ran
+          # 2-4x longer than the plain release targets, which had this all along.
+          restore-keys: |
+            cargo-windows-gui-${{ matrix.target }}-
+
+      - name: Download wintun
+        shell: pwsh
+        run: |
+          $ProgressPreference = 'SilentlyContinue'
+          
+          # Download wintun
+          New-Item -ItemType Directory -Force -Path "target/${{ matrix.target }}/release"
+          Invoke-WebRequest -Uri "https://www.wintun.net/builds/wintun-0.14.1.zip" -OutFile "target/wt.zip"
+          Expand-Archive "target/wt.zip" -DestinationPath "target/wt_tmp" -Force
+          Get-ChildItem "target/wt_tmp" -Filter "wintun.dll" -Recurse | Where-Object { $_.FullName -match 'bin[\\/]${{ matrix.arch }}[\\/]' } | Copy-Item -Destination "target/${{ matrix.target }}/release/wintun.dll" -Force
+
+      - name: Build Tauri App
+        working-directory: ostp-gui
+        run: |
+          npm install
+          cargo build -p ostp-tun-helper --release --target ${{ matrix.target }}
+          npx tauri build --no-bundle --target ${{ matrix.target }}
+
+      - name: Package Portable ZIP
+        shell: pwsh
+        run: |
+          $dir = "ostp-gui-dist"
+          New-Item -ItemType Directory -Force -Path $dir
+          Copy-Item "ostp-gui/src-tauri/target/${{ matrix.target }}/release/ostp-gui.exe" $dir
+          Copy-Item "target/${{ matrix.target }}/release/ostp-tun-helper.exe" $dir
+          Copy-Item "target/${{ matrix.target }}/release/wintun.dll" $dir
+          
+          Compress-Archive -Path "$dir/*" -DestinationPath "ostp-windows-gui-${{ matrix.arch }}.zip" -Force
+
+      - name: Upload to GitHub Release
+        uses: softprops/action-gh-release@v2
+        with:
+          # Computed once in resolve-channel so every platform/job in this run
+          # lands on the exact same tag: "{version}-alpha" / "{version}-beta"
+          # for rolling channel pushes, or the pushed "vX.Y.Z" tag as-is for a
+          # real stable release.
+          tag_name: ${{ needs.resolve-channel.outputs.tag_name }}
+          prerelease: ${{ needs.resolve-channel.outputs.prerelease }}
+          files: ostp-windows-gui-${{ matrix.arch }}.zip
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+
+  build-linux-gui:
+    name: Build Linux GUI (Tauri) - ${{ matrix.arch }}
+    needs: [check-and-test, resolve-channel]
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        include:
+          - arch: amd64
+            target: x86_64-unknown-linux-gnu
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 20
+
+      - name: Setup Rust toolchain
+        uses: dtolnay/rust-toolchain@stable
+        with:
+          targets: ${{ matrix.target }}
+
+      - name: Install Linux Dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y libwebkit2gtk-4.1-dev libappindicator3-dev librsvg2-dev patchelf
+
+      - name: Install Tauri CLI
+        run: npm install -g @tauri-apps/cli
+
+      - name: Cache cargo
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cargo/registry/index/
+            ~/.cargo/registry/cache/
+            ~/.cargo/git/db/
+            target/
+            ostp-gui/src-tauri/target/
+          key: cargo-linux-gui-${{ matrix.target }}-${{ hashFiles('**/Cargo.lock') }}
+          restore-keys: |
+            cargo-linux-gui-${{ matrix.target }}-
+
+      - name: Build Tauri App
+        working-directory: ostp-gui
+        run: |
+          npm install
+          # TUN mode shells out to this helper, elevated via pkexec. Only the
+          # Windows job used to build it, so the Linux package shipped without
+          # it and TUN could never start.
+          cargo build -p ostp-tun-helper --release --target ${{ matrix.target }} --manifest-path ../Cargo.toml
+          npx tauri build --no-bundle --target ${{ matrix.target }}
+
+      - name: Package Portable Tarball
+        run: |
+          set -euo pipefail
+          mkdir ostp-linux-gui-${{ matrix.arch }}
+          cp ostp-gui/src-tauri/target/${{ matrix.target }}/release/ostp-gui ostp-linux-gui-${{ matrix.arch }}/
+          # The GUI looks for the helper next to its own executable first.
+          cp target/${{ matrix.target }}/release/ostp-tun-helper ostp-linux-gui-${{ matrix.arch }}/
+          tar -czf ostp-linux-gui-${{ matrix.arch }}.tar.gz ostp-linux-gui-${{ matrix.arch }}
+
+      - name: Upload to GitHub Release
+        uses: softprops/action-gh-release@v2
+        with:
+          # Computed once in resolve-channel so every platform/job in this run
+          # lands on the exact same tag: "{version}-alpha" / "{version}-beta"
+          # for rolling channel pushes, or the pushed "vX.Y.Z" tag as-is for a
+          # real stable release.
+          tag_name: ${{ needs.resolve-channel.outputs.tag_name }}
+          prerelease: ${{ needs.resolve-channel.outputs.prerelease }}
+          files: ostp-linux-gui-${{ matrix.arch }}.tar.gz
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+
+  build-macos-gui:
+    name: Build macOS GUI (Tauri) - ${{ matrix.arch }}
+    needs: [check-and-test, resolve-channel]
+    runs-on: macos-latest
+    strategy:
+      matrix:
+        include:
+          - arch: amd64
+            target: x86_64-apple-darwin
+          - arch: arm64
+            target: aarch64-apple-darwin
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 20
+
+      - name: Setup Rust toolchain
+        uses: dtolnay/rust-toolchain@stable
+        with:
+          targets: ${{ matrix.target }}
+
+      - name: Install Tauri CLI
+        run: npm install -g @tauri-apps/cli
+
+      - name: Cache cargo
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cargo/registry/index/
+            ~/.cargo/registry/cache/
+            ~/.cargo/git/db/
+            target/
+            ostp-gui/src-tauri/target/
+          key: cargo-macos-gui-${{ matrix.target }}-${{ hashFiles('**/Cargo.lock') }}
+          restore-keys: |
+            cargo-macos-gui-${{ matrix.target }}-
+
+      - name: Build Tauri App
+        working-directory: ostp-gui
+        run: |
+          npm install
+          npx tauri build --no-bundle --target ${{ matrix.target }}
+
+      - name: Package Portable Tarball
+        run: |
+          mkdir ostp-macos-gui-${{ matrix.arch }}
+          cp ostp-gui/src-tauri/target/${{ matrix.target }}/release/ostp-gui ostp-macos-gui-${{ matrix.arch }}/
+          tar -czf ostp-macos-gui-${{ matrix.arch }}.tar.gz ostp-macos-gui-${{ matrix.arch }}
+
+      - name: Upload to GitHub Release
+        uses: softprops/action-gh-release@v2
+        with:
+          # Computed once in resolve-channel so every platform/job in this run
+          # lands on the exact same tag: "{version}-alpha" / "{version}-beta"
+          # for rolling channel pushes, or the pushed "vX.Y.Z" tag as-is for a
+          # real stable release.
+          tag_name: ${{ needs.resolve-channel.outputs.tag_name }}
+          prerelease: ${{ needs.resolve-channel.outputs.prerelease }}
+          files: ostp-macos-gui-${{ matrix.arch }}.tar.gz
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+
+  build-android:
+    name: Build Android Client (Flutter) - ${{ matrix.arch }}
+    needs: [check-and-test, resolve-channel]
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        include:
+          - arch: arm64-v8a
+            rust_target: aarch64-linux-android
+            flutter_target: android-arm64
+          - arch: armeabi-v7a
+            rust_target: armv7-linux-androideabi
+            flutter_target: android-arm
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Setup Java
+        uses: actions/setup-java@v3
+        with:
+          distribution: 'zulu'
+          java-version: '17'
+
+      - name: Setup Flutter
+        uses: subosito/flutter-action@v2
+        with:
+          flutter-version: '3.41.6'
+          channel: 'stable'
+
+      - name: Setup Rust toolchain
+        uses: dtolnay/rust-toolchain@stable
+        with:
+          targets: ${{ matrix.rust_target }}
+
+      - name: Setup Android NDK
+        uses: nttld/setup-ndk@v1
+        with:
+          ndk-version: r26b
+
+      # The Android jobs had no Rust caching at all, so every release recompiled
+      # the whole ostp-jni dependency graph from scratch — the main reason these
+      # were among the slowest jobs in the matrix.
+      - name: Cache cargo
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cargo/registry/index/
+            ~/.cargo/registry/cache/
+            ~/.cargo/git/db/
+            target/
+          key: cargo-android-${{ matrix.arch }}-${{ hashFiles('**/Cargo.lock') }}
+          restore-keys: |
+            cargo-android-${{ matrix.arch }}-
+
+      # cargo-ndk was built from source on every run. Cache the binary the same
+      # way the cross-compilation jobs already cache `cross`.
+      - name: Restore cargo-ndk binary cache
+        id: cargo-ndk-cache
+        uses: actions/cache@v4
+        with:
+          path: ~/.cargo/bin/cargo-ndk
+          key: cargo-ndk-bin-${{ runner.os }}-v1
+
+      - name: Install cargo-ndk (if not cached)
+        if: steps.cargo-ndk-cache.outputs.cache-hit != 'true'
+        run: cargo install cargo-ndk --locked
+
+      - name: Build Android APK
+        shell: bash
+        working-directory: ostp-flutter
+        env:
+          OSTP_KEYSTORE_B64: ${{ secrets.ANDROID_KEYSTORE_BASE64 }}
+          OSTP_KEYSTORE_PASSWORD: ${{ secrets.ANDROID_KEYSTORE_PASSWORD }}
+          OSTP_KEY_ALIAS: ${{ secrets.ANDROID_KEY_ALIAS }}
+          OSTP_KEY_PASSWORD: ${{ secrets.ANDROID_KEY_PASSWORD }}
+        run: |
+          set -euo pipefail
+
+          # 1. Materialise the upload keystore from secrets. Android keys an app
+          #    by applicationId + signing key and refuses to update across a key
+          #    change, so every published build MUST use this one key. Releases
+          #    used to fall through to the per-machine debug keystore, which on
+          #    ephemeral CI runners meant a different random key every build -
+          #    hence "App not installed" on upgrade.
+          if [ -z "${OSTP_KEYSTORE_B64:-}" ]; then
+            echo "::error::ANDROID_KEYSTORE_BASE64 secret is not set. Refusing to publish a"
+            echo "::error::debug-signed APK: users could not update over it and the key is"
+            echo "::error::not reproducible. See docs for the one-time keystore setup."
+            exit 1
+          fi
+          export OSTP_KEYSTORE_PATH="$RUNNER_TEMP/ostp-upload.jks"
+          # Strip any stray CR/LF before decoding: the secret is pasted from a
+          # shell whose line endings we don't control, and a single trailing \r
+          # is enough to corrupt the decode.
+          printf '%s' "$OSTP_KEYSTORE_B64" | tr -d '\r\n' | base64 -d > "$OSTP_KEYSTORE_PATH"
+
+          # Verify the keystore opens BEFORE spending four minutes on Gradle only
+          # to fail at the packaging step. The size/SHA-256 are safe to print (a
+          # hash reveals nothing) and let the operator compare against the local
+          # file to tell a transport problem apart from a wrong password.
+          echo "keystore: $(stat -c%s "$OSTP_KEYSTORE_PATH") bytes, sha256 $(sha256sum "$OSTP_KEYSTORE_PATH" | cut -d' ' -f1)"
+          if ! keytool -list -keystore "$OSTP_KEYSTORE_PATH" \
+                 -storepass "$OSTP_KEYSTORE_PASSWORD" >/dev/null 2>&1; then
+            echo "::error::The keystore did not open with ANDROID_KEYSTORE_PASSWORD."
+            echo "::error::If the SHA-256 above matches your local ostp-upload.jks, the file"
+            echo "::error::arrived intact and the password secret itself is wrong - note that"
+            echo "::error::PowerShell expands \$ inside double quotes, so a password containing"
+            echo "::error::one gets mangled unless it was set with single quotes."
+            exit 1
+          fi
+          if ! keytool -list -keystore "$OSTP_KEYSTORE_PATH" \
+                 -storepass "$OSTP_KEYSTORE_PASSWORD" -alias "$OSTP_KEY_ALIAS" >/dev/null 2>&1; then
+            echo "::error::Keystore opened, but it has no key under ANDROID_KEY_ALIAS."
+            echo "::error::Aliases present in the keystore:"
+            keytool -list -keystore "$OSTP_KEYSTORE_PATH" -storepass "$OSTP_KEYSTORE_PASSWORD" \
+              | grep -i "PrivateKeyEntry" || true
+            exit 1
+          fi
+
+          # 2. Compile JNI
+          mkdir -p android/app/src/main/jniLibs/${{ matrix.arch }}
+
+          cd ../ostp-jni
+          cargo ndk -t ${{ matrix.arch }} -o "../ostp-flutter/android/app/src/main/jniLibs" build --release
+          cd ../ostp-flutter
+
+          # 3. Build Flutter APK
+          flutter build apk --release --target-platform ${{ matrix.flutter_target }}
+
+          # 4. Fail loudly if the APK somehow still came out debug-signed, rather
+          #    than shipping another un-updatable build.
+          APK=build/app/outputs/flutter-apk/app-release.apk
+          if "$ANDROID_HOME"/build-tools/*/apksigner verify --print-certs "$APK" 2>/dev/null \
+               | grep -qi "CN=Android Debug"; then
+            echo "::error::APK is signed with the Android debug certificate - aborting."
+            exit 1
+          fi
+
+          # 5. Copy to output
+          cp "$APK" ostp-android-${{ matrix.arch }}.apk
+
+      - name: Upload to GitHub Release
+        uses: softprops/action-gh-release@v2
+        with:
+          # Computed once in resolve-channel so every platform/job in this run
+          # lands on the exact same tag: "{version}-alpha" / "{version}-beta"
+          # for rolling channel pushes, or the pushed "vX.Y.Z" tag as-is for a
+          # real stable release.
+          tag_name: ${{ needs.resolve-channel.outputs.tag_name }}
+          prerelease: ${{ needs.resolve-channel.outputs.prerelease }}
+          files: ostp-flutter/ostp-android-${{ matrix.arch }}.apk
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+
+

.gitignore

@@ -5,6 +5,7 @@
 **/*.rs.bk
 .idea/
 .vscode/
+**/node_modules/
 
 # Binaries & libraries
 *.exe
@@ -25,9 +26,34 @@ test_route.ps1
 config.json
 wintun.dll
 
+# Android signing keys. The upload keystore is the ONE key every published APK
+# must be signed with (Android refuses to update an app across a key change),
+# so losing or leaking it is unrecoverable — it can never be committed.
+*.jks
+*.keystore
+key.properties
+
+# Server runtime cache (public IP autodetect) — must never be committed,
+# it's regenerated locally and leaks whatever host it ran on last.
+.ostp_public_ip
+
 # Logs
 *.log
 
 # Dev notes (not for repo)
 .ai-rules.md
 turn-harvesting-idea.md
+
+# Private tooling (closed-source)
+ostp-prober/
+ostp-lab/
+
+ostp-brain/
+
+# Management panel built assets (built separately; dummy dist created for rust-embed build)
+ostp-control/
+
+.agents/
+netstack-smoltcp/
+dnstt/
+ostp-web/

.release-state.json

@@ -0,0 +1,6 @@
+{
+  "target_version": "0.4.4",
+  "branch": "master",
+  "alpha_iteration": 0,
+  "beta_iteration": 0
+}

CONTRIBUTING.md

@@ -0,0 +1,159 @@
+# Contributing to OSTP
+
+Thank you for your interest in contributing to **OSTP (Ospab Stealth Transport Protocol)**! We welcome contributions from developers, security researchers, testers, and documentation writers of all skill levels.
+
+By contributing to this project, you agree to abide by our code of conduct and license terms.
+
+---
+
+## Table of Contents
+
+1. [Development Setup](#development-setup)
+2. [Project Structure](#project-structure)
+3. [Branch Strategy](#branch-strategy)
+4. [Development Workflow](#development-workflow)
+5. [Commit Message Conventions](#commit-message-conventions)
+6. [Coding Guidelines](#coding-guidelines)
+7. [Submitting Pull Requests](#submitting-pull-requests)
+8. [Security Vulnerabilities](#security-vulnerabilities)
+
+---
+
+## Development Setup
+
+To build and test OSTP locally, you will need:
+
+*   **Rust Toolchain (1.75+)**: Install via [rustup](https://rustup.rs/).
+*   **Node.js (18+) & npm**: Required to build the frontend control panel (`ostp-control`) and compile Tauri GUI resources.
+*   **Git**: For version control.
+
+### Building the Project
+
+1.  **Clone the repository**:
+    ```bash
+    git clone https://github.com/ospab/ostp.git
+    cd ostp
+    ```
+
+2.  **Build the entire Cargo workspace**:
+    ```bash
+    cargo build
+    ```
+    `ostp-control` (the web panel) is only needed if you're working on it
+    specifically - the server build embeds a dummy `dist/` via `rust-embed`
+    otherwise, so this step is not required for day-to-day core/client/server
+    work. If you *are* touching the panel:
+    ```bash
+    cd ostp-control && npm install && npm run build && cd ..
+    ```
+
+3.  **Run tests**:
+    ```bash
+    cargo test --workspace
+    ```
+
+---
+
+## Project Structure
+
+The repository is organized as a Cargo workspace containing the following crates:
+
+*   [`ostp-core/`](file:///d:/ospab-projects/ostp/ostp-core): Core protocol logic, including packet formatting, serialization, selective ACK/NACK (ARQ) state machine, and the Noise protocol (`Noise_NNpsk0_25519_ChaChaPoly_BLAKE2s`) handshake.
+*   [`ostp-client/`](file:///d:/ospab-projects/ostp/ostp-client): Client implementations, including SOCKS5/HTTP local proxies, the native OSTP TUN interface routing, and split-tunneling bypass mechanisms.
+*   [`ostp-server/`](file:///d:/ospab-projects/ostp/ostp-server): Server logic, session dispatcher, anti-probing fallback server proxying, access key database, and the REST API for control panel communication.
+*   [`ostp-control/`](file:///d:/ospab-projects/ostp/ostp-control): A modern web dashboard for server administration (user management, real-time metrics, bandwidth limits).
+*   [`ostp-gui/`](file:///d:/ospab-projects/ostp/ostp-gui): Tauri-based desktop GUI application for Windows and Linux.
+*   [`ostp-flutter/`](file:///d:/ospab-projects/ostp/ostp-flutter): Mobile client code for Android platforms.
+
+---
+
+## Branch Strategy
+
+The repository runs three long-lived branches, in increasing order of stability:
+
+| Branch | Role |
+|---|---|
+| `alpha` | Active development. All feature work and fixes land here first. |
+| `beta` | Periodically fast-forwarded from `alpha` once it's had some soak time. Ships as the `{version}-beta` release channel. |
+| `master` | Fast-forwarded from `beta` when it's proven stable. Real, tagged releases (`vX.Y.Z`) are cut from here. |
+
+`beta` and `master` are **never** committed to directly - they only ever move forward by fast-forwarding from the branch below them. This means promotion is always a plain `git merge` with zero conflicts by construction: don't `git merge`/rebase feature work directly onto `beta` or `master`.
+
+**Contributor PRs target `alpha`**, not `master`.
+
+---
+
+## Development Workflow
+
+1.  **Check for existing issues** or open a new one to discuss proposed changes before starting work.
+2.  **Fork the repository** and create a new branch from `alpha`:
+    ```bash
+    git checkout alpha
+    git checkout -b feat/your-feature-name
+    ```
+3.  **Implement your changes**, ensuring you write appropriate unit or integration tests.
+4.  **Format your code**:
+    ```bash
+    cargo fmt --all
+    ```
+5.  **Run linter checks**:
+    ```bash
+    cargo clippy --workspace --all-targets -- -D warnings
+    ```
+6.  **Ensure all tests pass**:
+    ```bash
+    cargo test --workspace
+    ```
+
+---
+
+## Commit Message Conventions
+
+```
+<type>(<scope>): <short, imperative summary>
+
+<optional body - explain WHY, not what; the diff already shows what changed>
+```
+
+- **Type** - one of: `feat` (new capability), `fix` (bug fix), `docs`, `refactor` (no behavior change), `perf`, `test`, `chore` (deps/tooling/version bumps), `ci`, `security`.
+- **Scope** (optional) - the crate or area touched: `client`, `server`, `core`, `gui`, `flutter`, `ci`, `docs`, etc. e.g. `fix(client): ...`.
+- **Summary** - imperative mood ("add", not "added"/"adds"), no trailing period, ideally under ~70 characters.
+- **Body** - only when the *why* isn't obvious from the diff: a prior bug this fixes, a constraint that shaped the approach, a tradeoff you made. Don't restate what the diff already shows. Wrap at ~72 columns.
+
+```
+fix(server): drop junk frames by per-key marker instead of a global one
+
+A fixed 4-byte marker on every junk packet is itself a DPI signature any
+observer can filter on across every OSTP deployment. Derive the marker
+from the access key (HKDF, same scheme as obfuscation_key/psk) so it's
+per-user and indistinguishable from the packet's own random payload.
+```
+
+Multiple unrelated changes belong in separate commits, not one bundled commit - it keeps `git bisect` and review useful. Squash-merge is fine for a PR with a few "fix typo" / "address review" commits, but don't squash logically distinct changes together.
+
+---
+
+## Coding Guidelines
+
+*   **Safety**: Avoid using `unsafe` blocks unless absolutely necessary for low-level system bindings (e.g., FFI configurations like `setsockopt`). When using `unsafe`, add safety doc comments explaining why it is safe.
+*   **Documentation**: Document public modules, structs, and functions. Maintain comment integrity across codebase changes.
+*   **Logging**: Use the `tracing` framework for structured logging. Avoid `println!` for production logs.
+*   **Aesthetics**: When editing GUI or Web components, adhere to premium, modern web design aesthetics (vibrant color palettes, glassmorphism, responsive grids).
+
+---
+
+## Submitting Pull Requests
+
+1.  Push your branch to your GitHub fork:
+    ```bash
+    git push origin feat/your-feature-name
+    ```
+2.  Open a Pull Request (PR) targeting the `alpha` branch (see [Branch Strategy](#branch-strategy) - `master` only receives fast-forwards from `beta`, never direct PRs).
+3.  In your PR description, explain the rationale behind your changes, what was fixed/added, and how it was tested.
+4.  Verify that GitHub Actions CI runs successfully on your PR.
+
+---
+
+## Security Vulnerabilities
+
+If you discover a security-related vulnerability, please do **not** open a public issue. Instead, report it privately by emailing the core maintainers at [gvoprgrg@gmail.com](mailto:gvoprgrg@gmail.com). We will coordinate a swift disclosure and fix.

CONTRIBUTING.ru.md

@@ -0,0 +1,160 @@
+# Участие в разработке OSTP
+
+Спасибо за интерес к участию в разработке **OSTP (Ospab Stealth Transport Protocol)**! Мы рады любой помощи: от написания кода и тестирования до работы над документацией и проведения аудита безопасности.
+
+Присылая изменения в проект, вы соглашаетесь соблюдать правила нашего сообщества и условия лицензии.
+
+---
+
+## Содержание
+
+1. [Подготовка окружения](#подготовка-окружения)
+2. [Структура проекта](#структура-проекта)
+3. [Стратегия веток](#стратегия-веток)
+4. [Процесс разработки](#процесс-разработки)
+5. [Оформление коммитов](#оформление-коммитов)
+6. [Правила оформления кода](#правила-оформления-кода)
+7. [Создание Pull Request](#создание-pull-request)
+8. [Уязвимости безопасности](#уязвимости-безопасности)
+
+---
+
+## Подготовка окружения
+
+Для локальной сборки и тестирования OSTP вам понадобятся:
+
+*   **Rust Toolchain (1.75+)**: Рекомендуется установить через [rustup](https://rustup.rs/).
+*   **Node.js (18+) и npm**: Необходимы для сборки веб-панели управления (`ostp-control`) и сборки интерфейса Tauri.
+*   **Git**: Для контроля версий.
+
+### Сборка проекта
+
+1.  **Клонируйте репозиторий**:
+    ```bash
+    git clone https://github.com/ospab/ostp.git
+    cd ostp
+    ```
+
+2.  **Соберите весь Cargo-workspace**:
+    ```bash
+    cargo build
+    ```
+    `ostp-control` (веб-панель) нужна только если вы работаете конкретно над
+    ней - в остальных случаях сервер собирается с пустым `dist/` через
+    `rust-embed`, и этот шаг не нужен для повседневной работы над
+    core/client/server. Если вы всё же трогаете панель:
+    ```bash
+    cd ostp-control && npm install && npm run build && cd ..
+    ```
+
+3.  **Запустите тесты**:
+    ```bash
+    cargo test --workspace
+    ```
+
+---
+
+## Структура проекта
+
+Репозиторий представляет собой единый Cargo-workspace со следующими компонентами:
+
+*   [`ostp-core/`](file:///d:/ospab-projects/ostp/ostp-core): Базовая логика протокола: форматирование пакетов, сериализация, конечный автомат выборочного подтверждения (ARQ/ACK/NACK) и рукопожатие Noise (`Noise_NNpsk0_25519_ChaChaPoly_BLAKE2s`).
+*   [`ostp-client/`](file:///d:/ospab-projects/ostp/ostp-client): Клиентская часть: локальные SOCKS5/HTTP прокси-серверы, нативный OSTP TUN-интерфейс (через драйвер `wintun`) и реализация раздельного туннелирования для прямого обхода трафика.
+*   [`ostp-server/`](file:///d:/ospab-projects/ostp/ostp-server): Серверная часть: диспетчеризация сессий, маскировка под классические веб-серверы при активном сканировании, база данных ключей доступа и REST API панели управления.
+*   [`ostp-control/`](file:///d:/ospab-projects/ostp/ostp-control): Панель администратора (пользователи, статистика трафика в реальном времени, лимиты скорости и объема данных).
+*   [`ostp-gui/`](file:///d:/ospab-projects/ostp/ostp-gui): Настольное приложение-клиент для Windows и Linux на платформе Tauri.
+*   [`ostp-flutter/`](file:///d:/ospab-projects/ostp/ostp-flutter): Мобильный клиент для платформы Android.
+
+---
+
+## Стратегия веток
+
+В репозитории три долгоживущие ветки, по возрастанию стабильности:
+
+| Ветка | Роль |
+|---|---|
+| `alpha` | Активная разработка. Вся новая работа и фиксы попадают сюда первыми. |
+| `beta` | Периодически перематывается вперёд (fast-forward) от `alpha`, когда та немного «отлежалась». Собирается в канал релиза `{версия}-beta`. |
+| `master` | Перематывается вперёд от `beta`, когда та доказала стабильность. Настоящие тегированные релизы (`vX.Y.Z`) режутся отсюда. |
+
+В `beta` и `master` **никогда** не коммитят напрямую - они только перематываются вперёд от ветки уровнем ниже. Это значит, что промоушен - всегда обычный `git merge` без единого конфликта по построению: не мержите/не ребейзьте свою фичу прямо в `beta` или `master`.
+
+**PR от контрибьюторов нацелены на `alpha`**, не на `master`.
+
+---
+
+## Процесс разработки
+
+1.  **Проверьте существующие задачи** или откройте новую тему (Issue) для обсуждения предлагаемых изменений.
+2.  **Сделайте fork репозитория** и создайте новую ветку от `alpha`:
+    ```bash
+    git checkout alpha
+    git checkout -b feat/имя-вашей-фичи
+    ```
+3.  **Внесите необходимые изменения** и добавьте соответствующие модульные или интеграционные тесты.
+4.  **Выровняйте форматирование кода**:
+    ```bash
+    cargo fmt --all
+    ```
+5.  **Запустите статический анализатор**:
+    ```bash
+    cargo clippy --workspace --all-targets -- -D warnings
+    ```
+6.  **Убедитесь, что все тесты проходят**:
+    ```bash
+    cargo test --workspace
+    ```
+
+---
+
+## Оформление коммитов
+
+```
+<тип>(<область>): <краткое описание в повелительном наклонении>
+
+<опционально: тело - объясняет ПОЧЕМУ, а не что; диф и так показывает что изменилось>
+```
+
+- **Тип** - один из: `feat` (новая функциональность), `fix` (исправление бага), `docs`, `refactor` (без изменения поведения), `perf`, `test`, `chore` (зависимости/тулинг/версии), `ci`, `security`.
+- **Область** (опционально) - крейт или часть проекта: `client`, `server`, `core`, `gui`, `flutter`, `ci`, `docs` и т.д., например `fix(client): ...`.
+- **Краткое описание** - повелительное наклонение ("добавь", а не "добавил"/"добавляет"), без точки в конце, желательно до ~70 символов.
+- **Тело** - только когда причина не очевидна из дифа: какой баг это чинит, какое ограничение определило подход, на какой trade-off вы пошли. Не пересказывайте то, что и так видно в дифе. Перенос строк на ~72 символах.
+
+```
+fix(server): отбрасывать junk-фреймы по маркеру для каждого ключа, а не глобальному
+
+Фиксированный 4-байтовый маркер на каждом junk-пакете сам по себе - сигнатура
+DPI, по которой можно фильтровать любого наблюдателя во всех деплойментах OSTP
+сразу. Выводим маркер из access_key (HKDF, та же схема что у
+obfuscation_key/psk), чтобы он был индивидуальным для ключа и неотличимым от
+случайной полезной нагрузки пакета.
+```
+
+Несколько несвязанных изменений - это несколько отдельных коммитов, а не один сборный. Это сохраняет пользу от `git bisect` и код-ревью. Squash-merge подходит для PR с парой коммитов вроде "fix typo" / "address review", но не сквошьте вместе логически разные изменения.
+
+---
+
+## Правила оформления кода
+
+*   **Безопасность (Safety)**: Избегайте использования блоков `unsafe` везде, где это возможно. Допускается их использование только для низкоуровневых системных вызовов (например, FFI-настройки сокетов `setsockopt`). Любой блок `unsafe` должен сопровождаться комментарием `// SAFETY: ...`.
+*   **Документация**: Пишите документацию для публичных модулей, структур и методов. Сохраняйте целостность комментариев при рефакторинге.
+*   **Логирование**: Используйте фреймворк `tracing` для структурированного логирования. Не используйте `println!` в рабочем коде.
+*   **Дизайн**: При изменении веб-интерфейсов или GUI следуйте современным визуальным трендам (плавные анимации, сбалансированная цветовая гамма, адаптивная верстка).
+
+---
+
+## Создание Pull Request
+
+1.  Отправьте ветку в ваш fork-репозиторий:
+    ```bash
+    git push origin feat/имя-вашей-фичи
+    ```
+2.  Создайте Pull Request (PR) в ветку `alpha` основного репозитория (см. [Стратегия веток](#стратегия-веток) - `master` получает только fast-forward от `beta`, PR туда не принимаются напрямую).
+3.  Подробно опишите внесенные изменения: какая проблема решается, как проводилось тестирование и на каких платформах проверялась сборка.
+4.  Убедитесь, что автоматическое тестирование (GitHub Actions CI) завершилось успешно.
+
+---
+
+## Уязвимости безопасности
+
+Если вы обнаружили уязвимость, пожалуйста, **не** публикуйте её в открытых Issue. Вместо этого отправьте отчёт разработчикам на почту [gvoprgrg@gmail.com](mailto:gvoprgrg@gmail.com) для координации закрытого исправления.

Cargo.toml

@@ -4,26 +4,25 @@ members = [
     "ostp-client",
     "ostp-server",
     "ostp-jni", "ostp",
-]
+    "ostp-tun-helper"
+, "ostp-tun"]
+exclude = ["ostp-gui/src-tauri", "ostp-brain", "ostp-prober"]
 resolver = "2"
 
 [workspace.package]
 edition = "2021"
-license = "BSL 1.1"
-version = "0.1.37"
+license = "AGPL-3.0"
+version = "0.4.4"
 
 [workspace.dependencies]
 anyhow = "1.0"
-async-trait = "0.1"
 bytes = "1.6"
 chacha20poly1305 = "0.10"
 rand = "0.8"
-rand_distr = "0.4"
-snow = "0.9"
+snow = { version = "0.9", features = ["risky-raw-split"] }
 thiserror = "1.0"
 tokio = { version = "1.37", features = ["rt-multi-thread", "macros", "net", "time", "io-util", "sync", "signal"] }
 tracing = "0.1"
-x25519-dalek = "2"
 sha2 = "0.10"
 hmac = "0.12"
 portable-atomic = "1.10"

LICENSE

@@ -1,74 +1,661 @@
-Business Source License 1.1
+                    GNU AFFERO GENERAL PUBLIC LICENSE
+                       Version 3, 19 November 2007
 
-Parameters
+ Copyright (C) 2007 Free Software Foundation, Inc. <https://fsf.org/>
+ Everyone is permitted to copy and distribute verbatim copies
+ of this license document, but changing it is not allowed.
 
-Licensor:                     Ospab Foundation (represented by Syralev Georgiy)
-Licensed Work:                The Ospab Stealth Transport Protocol (OSTP) and all
-                              associated workspace crates, utilities, and documents.
-Additional Use Grant:         The Licensor hereby grants you the right to copy,
-                              modify, create derivative works, redistribute, and
-                              make non-production and non-commercial use of the
-                              Licensed Work. You are also permitted to use the
-                              Licensed Work in production for personal, private
-                              utility and non-profit organizations.
-Change Date:                  May 14, 2030
-Change License:               MIT License (as defined below)
+                            Preamble
 
------------------------------------------------------------------------------------
+  The GNU Affero General Public License is a free, copyleft license for
+software and other kinds of works, specifically designed to ensure
+cooperation with the community in the case of network server software.
 
-Terms
+  The licenses for most software and other practical works are designed
+to take away your freedom to share and change the works.  By contrast,
+our General Public Licenses are intended to guarantee your freedom to
+share and change all versions of a program--to make sure it remains free
+software for all its users.
 
-1. The Licensor hereby grants you the right to copy, modify, create derivative works, 
-   redistribute, and make use of the Licensed Work only as permitted by the 
-   Additional Use Grant.
+  When we speak of free software, we are referring to freedom, not
+price.  Our General Public Licenses are designed to make sure that you
+have the freedom to distribute copies of free software (and charge for
+them if you wish), that you receive source code or can get it if you
+want it, that you can change the software or use pieces of it in new
+free programs, and that you know you can do these things.
 
-2. The Licensor hereby grants you the right to copy, modify, create derivative works, 
-   redistribute, and make use of the Licensed Work under the terms of the Change 
-   License on and after the Change Date.
+  Developers that use our General Public Licenses protect your rights
+with two steps: (1) assert copyright on the software, and (2) offer
+you this License which gives you legal permission to copy, distribute
+and/or modify the software.
 
-3. To the extent that any term of this License (including the Additional Use Grant 
-   and the Change License) is in conflict with the Terms of this License, these 
-   Terms shall take precedence.
+  A secondary benefit of defending all users' freedom is that
+improvements made in alternate versions of the program, if they
+receive widespread use, become available for other developers to
+incorporate.  Many developers of free software are heartened and
+encouraged by the resulting cooperation.  However, in the case of
+software used on network servers, this result may fail to come about.
+The GNU General Public License permits making a modified version and
+letting the public access it on a server without ever releasing its
+source code to the public.
 
-4. Every copy of the Licensed Work and any derivative work must include this 
-   License and all other copyright, trademark, and proprietary notices included 
-   with the Licensed Work.
+  The GNU Affero General Public License is designed specifically to
+ensure that, in such cases, the modified source code becomes available
+to the community.  It requires the operator of a network server to
+provide the source code of the modified version running there to the
+users of that server.  Therefore, public use of a modified version, on
+a publicly accessible server, gives the public access to the source
+code of the modified version.
 
-5. Any use of the Licensed Work that is not permitted by this License is a breach 
-   of this License and may terminate your rights under this License.
+  An older license, called the Affero General Public License and
+published by Affero, was designed to accomplish similar goals.  This is
+a different license, not a version of the Affero GPL, but Affero has
+released a new version of the Affero GPL which permits relicensing under
+this license.
 
-6. DISCLAIMER OF WARRANTY. TO THE EXTENT PERMITTED BY APPLICABLE LAW, THE LICENSED 
-   WORK IS PROVIDED ON AN "AS IS" BASIS. THE LICENSOR MAKES NO REPRESENTATIONS OR 
-   WARRANTIES OF ANY KIND CONCERNING THE LICENSED WORK, EXPRESS OR IMPLIED, STATUTORY 
-   OR OTHERWISE, INCLUDING, WITHOUT LIMITATION, WARRANTIES OF TITLE, 
-   MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, OR NONINFRINGEMENT.
+  The precise terms and conditions for copying, distribution and
+modification follow.
 
-7. LIMITATION OF LIABILITY. TO THE EXTENT PERMITTED BY APPLICABLE LAW, IN NO EVENT 
-   WILL THE LICENSOR BE LIABLE TO YOU ON ANY LEGAL THEORY FOR ANY SPECIAL, INCIDENTAL, 
-   CONSEQUENTIAL, PUNITIVE, OR EXEMPLARY DAMAGES ARISING OUT OF THIS LICENSE OR THE 
-   USE OF THE LICENSED WORK, EVEN IF THE LICENSOR HAS BEEN ADVISED OF THE 
-   POSSIBILITY OF SUCH DAMAGES.
+                       TERMS AND CONDITIONS
 
------------------------------------------------------------------------------------
+  0. Definitions.
 
-Change License Text (MIT License)
+  "This License" refers to version 3 of the GNU Affero General Public License.
 
-Copyright (c) 2026 Syralev Georgiy (Ospab Foundation)
+  "Copyright" also means copyright-like laws that apply to other kinds of
+works, such as semiconductor masks.
 
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
+  "The Program" refers to any copyrightable work licensed under this
+License.  Each licensee is addressed as "you".  "Licensees" and
+"recipients" may be individuals or organizations.
 
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
+  To "modify" a work means to copy from or adapt all or part of the work
+in a fashion requiring copyright permission, other than the making of an
+exact copy.  The resulting work is called a "modified version" of the
+earlier work or a work "based on" the earlier work.
 
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
+  A "covered work" means either the unmodified Program or a work based
+on the Program.
+
+  To "propagate" a work means to do anything with it that, without
+permission, would make you directly or secondarily liable for
+infringement under applicable copyright law, except executing it on a
+computer or modifying a private copy.  Propagation includes copying,
+distribution (with or without modification), making available to the
+public, and in some countries other activities as well.
+
+  To "convey" a work means any kind of propagation that enables other
+parties to make or receive copies.  Mere interaction with a user through
+a computer network, with no transfer of a copy, is not conveying.
+
+  An interactive user interface displays "Appropriate Legal Notices"
+to the extent that it includes a convenient and prominently visible
+feature that (1) displays an appropriate copyright notice, and (2)
+tells the user that there is no warranty for the work (except to the
+extent that warranties are provided), that licensees may convey the
+work under this License, and how to view a copy of this License.  If
+the interface presents a list of user commands or options, such as a
+menu, a prominent item in the list meets this criterion.
+
+  1. Source Code.
+
+  The "source code" for a work means the preferred form of the work
+for making modifications to it.  "Object code" means any non-source
+form of a work.
+
+  A "Standard Interface" means an interface that either is an official
+standard defined by a recognized standards body, or, in the case of
+interfaces specified for a particular programming language, one that
+is widely used among developers working in that language.
+
+  The "System Libraries" of an executable work include anything, other
+than the work as a whole, that (a) is included in the normal form of
+packaging a Major Component, but which is not part of that Major
+Component, and (b) serves only to enable use of the work with that
+Major Component, or to implement a Standard Interface for which an
+implementation is available to the public in source code form.  A
+"Major Component", in this context, means a major essential component
+(kernel, window system, and so on) of the specific operating system
+(if any) on which the executable work runs, or a compiler used to
+produce the work, or an object code interpreter used to run it.
+
+  The "Corresponding Source" for a work in object code form means all
+the source code needed to generate, install, and (for an executable
+work) run the object code and to modify the work, including scripts to
+control those activities.  However, it does not include the work's
+System Libraries, or general-purpose tools or generally available free
+programs which are used unmodified in performing those activities but
+which are not part of the work.  For example, Corresponding Source
+includes interface definition files associated with source files for
+the work, and the source code for shared libraries and dynamically
+linked subprograms that the work is specifically designed to require,
+such as by intimate data communication or control flow between those
+subprograms and other parts of the work.
+
+  The Corresponding Source need not include anything that users
+can regenerate automatically from other parts of the Corresponding
+Source.
+
+  The Corresponding Source for a work in source code form is that
+same work.
+
+  2. Basic Permissions.
+
+  All rights granted under this License are granted for the term of
+copyright on the Program, and are irrevocable provided the stated
+conditions are met.  This License explicitly affirms your unlimited
+permission to run the unmodified Program.  The output from running a
+covered work is covered by this License only if the output, given its
+content, constitutes a covered work.  This License acknowledges your
+rights of fair use or other equivalent, as provided by copyright law.
+
+  You may make, run and propagate covered works that you do not
+convey, without conditions so long as your license otherwise remains
+in force.  You may convey covered works to others for the sole purpose
+of having them make modifications exclusively for you, or provide you
+with facilities for running those works, provided that you comply with
+the terms of this License in conveying all material for which you do
+not control copyright.  Those thus making or running the covered works
+for you must do so exclusively on your behalf, under your direction
+and control, on terms that prohibit them from making any copies of
+your copyrighted material outside their relationship with you.
+
+  Conveying under any other circumstances is permitted solely under
+the conditions stated below.  Sublicensing is not allowed; section 10
+makes it unnecessary.
+
+  3. Protecting Users' Legal Rights From Anti-Circumvention Law.
+
+  No covered work shall be deemed part of an effective technological
+measure under any applicable law fulfilling obligations under article
+11 of the WIPO copyright treaty adopted on 20 December 1996, or
+similar laws prohibiting or restricting circumvention of such
+measures.
+
+  When you convey a covered work, you waive any legal power to forbid
+circumvention of technological measures to the extent such circumvention
+is effected by exercising rights under this License with respect to
+the covered work, and you disclaim any intention to limit operation or
+modification of the work as a means of enforcing, against the work's
+users, your or third parties' legal rights to forbid circumvention of
+technological measures.
+
+  4. Conveying Verbatim Copies.
+
+  You may convey verbatim copies of the Program's source code as you
+receive it, in any medium, provided that you conspicuously and
+appropriately publish on each copy an appropriate copyright notice;
+keep intact all notices stating that this License and any
+non-permissive terms added in accord with section 7 apply to the code;
+keep intact all notices of the absence of any warranty; and give all
+recipients a copy of this License along with the Program.
+
+  You may charge any price or no price for each copy that you convey,
+and you may offer support or warranty protection for a fee.
+
+  5. Conveying Modified Source Versions.
+
+  You may convey a work based on the Program, or the modifications to
+produce it from the Program, in the form of source code under the
+terms of section 4, provided that you also meet all of these conditions:
+
+    a) The work must carry prominent notices stating that you modified
+    it, and giving a relevant date.
+
+    b) The work must carry prominent notices stating that it is
+    released under this License and any conditions added under section
+    7.  This requirement modifies the requirement in section 4 to
+    "keep intact all notices".
+
+    c) You must license the entire work, as a whole, under this
+    License to anyone who comes into possession of a copy.  This
+    License will therefore apply, along with any applicable section 7
+    additional terms, to the whole of the work, and all its parts,
+    regardless of how they are packaged.  This License gives no
+    permission to license the work in any other way, but it does not
+    invalidate such permission if you have separately received it.
+
+    d) If the work has interactive user interfaces, each must display
+    Appropriate Legal Notices; however, if the Program has interactive
+    interfaces that do not display Appropriate Legal Notices, your
+    work need not make them do so.
+
+  A compilation of a covered work with other separate and independent
+works, which are not by their nature extensions of the covered work,
+and which are not combined with it such as to form a larger program,
+in or on a volume of a storage or distribution medium, is called an
+"aggregate" if the compilation and its resulting copyright are not
+used to limit the access or legal rights of the compilation's users
+beyond what the individual works permit.  Inclusion of a covered work
+in an aggregate does not cause this License to apply to the other
+parts of the aggregate.
+
+  6. Conveying Non-Source Forms.
+
+  You may convey a covered work in object code form under the terms
+of sections 4 and 5, provided that you also convey the
+machine-readable Corresponding Source under the terms of this License,
+in one of these ways:
+
+    a) Convey the object code in, or embodied in, a physical product
+    (including a physical distribution medium), accompanied by the
+    Corresponding Source fixed on a durable physical medium
+    customarily used for software interchange.
+
+    b) Convey the object code in, or embodied in, a physical product
+    (including a physical distribution medium), accompanied by a
+    written offer, valid for at least three years and valid for as
+    long as you offer spare parts or customer support for that product
+    model, to give anyone who possesses the object code either (1) a
+    copy of the Corresponding Source for all the software in the
+    product that is covered by this License, on a durable physical
+    medium customarily used for software interchange, for a price no
+    more than your reasonable cost of physically performing this
+    conveying of source, or (2) access to copy the
+    Corresponding Source from a network server at no charge.
+
+    c) Convey individual copies of the object code with a copy of the
+    written offer to provide the Corresponding Source.  This
+    alternative is allowed only occasionally and noncommercially, and
+    only if you received the object code with such an offer, in accord
+    with subsection 6b.
+
+    d) Convey the object code by offering access from a designated
+    place (gratis or for a charge), and offer equivalent access to the
+    Corresponding Source in the same way through the same place at no
+    further charge.  You need not require recipients to copy the
+    Corresponding Source along with the object code.  If the place to
+    copy the object code is a network server, the Corresponding Source
+    may be on a different server (operated by you or a third party)
+    that supports equivalent copying facilities, provided you maintain
+    clear directions next to the object code saying where to find the
+    Corresponding Source.  Regardless of what server hosts the
+    Corresponding Source, you remain obligated to ensure that it is
+    available for as long as needed to satisfy these requirements.
+
+    e) Convey the object code using peer-to-peer transmission, provided
+    you inform other peers where the object code and Corresponding
+    Source of the work are being offered to the general public at no
+    charge under subsection 6d.
+
+  A separable portion of the object code, whose source code is excluded
+from the Corresponding Source as a System Library, need not be
+included in conveying the object code work.
+
+  A "User Product" is either (1) a "consumer product", which means any
+tangible personal property which is normally used for personal, family,
+or household purposes, or (2) anything designed or sold for incorporation
+into a dwelling.  In determining whether a product is a consumer product,
+doubtful cases shall be resolved in favor of coverage.  For a particular
+product received by a particular user, "normally used" refers to a
+typical or common use of that class of product, regardless of the status
+of the particular user or of the way in which the particular user
+actually uses, or expects or is expected to use, the product.  A product
+is a consumer product regardless of whether the product has substantial
+commercial, industrial or non-consumer uses, unless such uses represent
+the only significant mode of use of the product.
+
+  "Installation Information" for a User Product means any methods,
+procedures, authorization keys, or other information required to install
+and execute modified versions of a covered work in that User Product from
+a modified version of its Corresponding Source.  The information must
+suffice to ensure that the continued functioning of the modified object
+code is in no case prevented or interfered with solely because
+modification has been made.
+
+  If you convey an object code work under this section in, or with, or
+specifically for use in, a User Product, and the conveying occurs as
+part of a transaction in which the right of possession and use of the
+User Product is transferred to the recipient in perpetuity or for a
+fixed term (regardless of how the transaction is characterized), the
+Corresponding Source conveyed under this section must be accompanied
+by the Installation Information.  But this requirement does not apply
+if neither you nor any third party retains the ability to install
+modified object code on the User Product (for example, the work has
+been installed in ROM).
+
+  The requirement to provide Installation Information does not include a
+requirement to continue to provide support service, warranty, or updates
+for a work that has been modified or installed by the recipient, or for
+the User Product in which it has been modified or installed.  Access to a
+network may be denied when the modification itself materially and
+adversely affects the operation of the network or violates the rules and
+protocols for communication across the network.
+
+  Corresponding Source conveyed, and Installation Information provided,
+in accord with this section must be in a format that is publicly
+documented (and with an implementation available to the public in
+source code form), and must require no special password or key for
+unpacking, reading or copying.
+
+  7. Additional Terms.
+
+  "Additional permissions" are terms that supplement the terms of this
+License by making exceptions from one or more of its conditions.
+Additional permissions that are applicable to the entire Program shall
+be treated as though they were included in this License, to the extent
+that they are valid under applicable law.  If additional permissions
+apply only to part of the Program, that part may be used separately
+under those permissions, but the entire Program remains governed by
+this License without regard to the additional permissions.
+
+  When you convey a copy of a covered work, you may at your option
+remove any additional permissions from that copy, or from any part of
+it.  (Additional permissions may be written to require their own
+removal in certain cases when you modify the work.)  You may place
+additional permissions on material, added by you to a covered work,
+for which you have or can give appropriate copyright permission.
+
+  Notwithstanding any other provision of this License, for material you
+add to a covered work, you may (if authorized by the copyright holders of
+that material) supplement the terms of this License with terms:
+
+    a) Disclaiming warranty or limiting liability differently from the
+    terms of sections 15 and 16 of this License; or
+
+    b) Requiring preservation of specified reasonable legal notices or
+    author attributions in that material or in the Appropriate Legal
+    Notices displayed by works containing it; or
+
+    c) Prohibiting misrepresentation of the origin of that material, or
+    requiring that modified versions of such material be marked in
+    reasonable ways as different from the original version; or
+
+    d) Limiting the use for publicity purposes of names of licensors or
+    authors of the material; or
+
+    e) Declining to grant rights under trademark law for use of some
+    trade names, trademarks, or service marks; or
+
+    f) Requiring indemnification of licensors and authors of that
+    material by anyone who conveys the material (or modified versions of
+    it) with contractual assumptions of liability to the recipient, for
+    any liability that these contractual assumptions directly impose on
+    those licensors and authors.
+
+  All other non-permissive additional terms are considered "further
+restrictions" within the meaning of section 10.  If the Program as you
+received it, or any part of it, contains a notice stating that it is
+governed by this License along with a term that is a further
+restriction, you may remove that term.  If a license document contains
+a further restriction but permits relicensing or conveying under this
+License, you may add to a covered work material governed by the terms
+of that license document, provided that the further restriction does
+not survive such relicensing or conveying.
+
+  If you add terms to a covered work in accord with this section, you
+must place, in the relevant source files, a statement of the
+additional terms that apply to those files, or a notice indicating
+where to find the applicable terms.
+
+  Additional terms, permissive or non-permissive, may be stated in the
+form of a separately written license, or stated as exceptions;
+the above requirements apply either way.
+
+  8. Termination.
+
+  You may not propagate or modify a covered work except as expressly
+provided under this License.  Any attempt otherwise to propagate or
+modify it is void, and will automatically terminate your rights under
+this License (including any patent licenses granted under the third
+paragraph of section 11).
+
+  However, if you cease all violation of this License, then your
+license from a particular copyright holder is reinstated (a)
+provisionally, unless and until the copyright holder explicitly and
+finally terminates your license, and (b) permanently, if the copyright
+holder fails to notify you of the violation by some reasonable means
+prior to 60 days after the cessation.
+
+  Moreover, your license from a particular copyright holder is
+reinstated permanently if the copyright holder notifies you of the
+violation by some reasonable means, this is the first time you have
+received notice of violation of this License (for any work) from that
+copyright holder, and you cure the violation prior to 30 days after
+your receipt of the notice.
+
+  Termination of your rights under this section does not terminate the
+licenses of parties who have received copies or rights from you under
+this License.  If your rights have been terminated and not permanently
+reinstated, you do not qualify to receive new licenses for the same
+material under section 10.
+
+  9. Acceptance Not Required for Having Copies.
+
+  You are not required to accept this License in order to receive or
+run a copy of the Program.  Ancillary propagation of a covered work
+occurring solely as a consequence of using peer-to-peer transmission
+to receive a copy likewise does not require acceptance.  However,
+nothing other than this License grants you permission to propagate or
+modify any covered work.  These actions infringe copyright if you do
+not accept this License.  Therefore, by modifying or propagating a
+covered work, you indicate your acceptance of this License to do so.
+
+  10. Automatic Licensing of Downstream Recipients.
+
+  Each time you convey a covered work, the recipient automatically
+receives a license from the original licensors, to run, modify and
+propagate that work, subject to this License.  You are not responsible
+for enforcing compliance by third parties with this License.
+
+  An "entity transaction" is a transaction transferring control of an
+organization, or substantially all assets of one, or subdividing an
+organization, or merging organizations.  If propagation of a covered
+work results from an entity transaction, each party to that
+transaction who receives a copy of the work also receives whatever
+licenses to the work the party's predecessor in interest had or could
+give under the previous paragraph, plus a right to possession of the
+Corresponding Source of the work from the predecessor in interest, if
+the predecessor has it or can get it with reasonable efforts.
+
+  You may not impose any further restrictions on the exercise of the
+rights granted or affirmed under this License.  For example, you may
+not impose a license fee, royalty, or other charge for exercise of
+rights granted under this License, and you may not initiate litigation
+(including a cross-claim or counterclaim in a lawsuit) alleging that
+any patent claim is infringed by making, using, selling, offering for
+sale, or importing the Program or any portion of it.
+
+  11. Patents.
+
+  A "contributor" is a copyright holder who authorizes use under this
+License of the Program or a work on which the Program is based.  The
+work thus licensed is called the contributor's "contributor version".
+
+  A contributor's "essential patent claims" are all patent claims
+owned or controlled by the contributor, whether already acquired or
+hereafter acquired, that would be infringed by some manner, permitted
+by this License, of making, using, or selling its contributor version,
+but do not include claims that would be infringed only as a
+consequence of further modification of the contributor version.  For
+purposes of this definition, "control" includes the right to grant
+patent sublicenses in a manner consistent with the requirements of
+this License.
+
+  Each contributor grants you a non-exclusive, worldwide, royalty-free
+patent license under the contributor's essential patent claims, to
+make, use, sell, offer for sale, import and otherwise run, modify and
+propagate the contents of its contributor version.
+
+  In the following three paragraphs, a "patent license" is any express
+agreement or commitment, however denominated, not to enforce a patent
+(such as an express permission to practice a patent or covenant not to
+sue for patent infringement).  To "grant" such a patent license to a
+party means to make such an agreement or commitment not to enforce a
+patent against the party.
+
+  If you convey a covered work, knowingly relying on a patent license,
+and the Corresponding Source of the work is not available for anyone
+to copy, free of charge and under the terms of this License, through a
+publicly available network server or other readily accessible means,
+then you must either (1) cause the Corresponding Source to be so
+available, or (2) arrange to deprive yourself of the benefit of the
+patent license for this particular work, or (3) arrange, in a manner
+consistent with the requirements of this License, to extend the patent
+license to downstream recipients.  "Knowingly relying" means you have
+actual knowledge that, but for the patent license, your conveying the
+covered work in a country, or your recipient's use of the covered work
+in a country, would infringe one or more identifiable patents in that
+country that you have reason to believe are valid.
+
+  If, pursuant to or in connection with a single transaction or
+arrangement, you convey, or propagate by procuring conveyance of, a
+covered work, and grant a patent license to some of the parties
+receiving the covered work authorizing them to use, propagate, modify
+or convey a specific copy of the covered work, then the patent license
+you grant is automatically extended to all recipients of the covered
+work and works based on it.
+
+  A patent license is "discriminatory" if it does not include within
+the scope of its coverage, prohibits the exercise of, or is
+conditioned on the non-exercise of one or more of the rights that are
+specifically granted under this License.  You may not convey a covered
+work if you are a party to an arrangement with a third party that is
+in the business of distributing software, under which you make payment
+to the third party based on the extent of your activity of conveying
+the work, and under which the third party grants, to any of the
+parties who would receive the covered work from you, a discriminatory
+patent license (a) in connection with copies of the covered work
+conveyed by you (or copies made from those copies), or (b) primarily
+for and in connection with specific products or compilations that
+contain the covered work, unless you entered into that arrangement,
+or that patent license was granted, prior to 28 March 2007.
+
+  Nothing in this License shall be construed as excluding or limiting
+any implied license or other defenses to infringement that may
+otherwise be available to you under applicable patent law.
+
+  12. No Surrender of Others' Freedom.
+
+  If conditions are imposed on you (whether by court order, agreement or
+otherwise) that contradict the conditions of this License, they do not
+excuse you from the conditions of this License.  If you cannot convey a
+covered work so as to satisfy simultaneously your obligations under this
+License and any other pertinent obligations, then as a consequence you may
+not convey it at all.  For example, if you agree to terms that obligate you
+to collect a royalty for further conveying from those to whom you convey
+the Program, the only way you could satisfy both those terms and this
+License would be to refrain entirely from conveying the Program.
+
+  13. Remote Network Interaction; Use with the GNU General Public License.
+
+  Notwithstanding any other provision of this License, if you modify the
+Program, your modified version must prominently offer all users
+interacting with it remotely through a computer network (if your version
+supports such interaction) an opportunity to receive the Corresponding
+Source of your version by providing access to the Corresponding Source
+from a network server at no charge, through some standard or customary
+means of facilitating copying of software.  This Corresponding Source
+shall include the Corresponding Source for any work covered by version 3
+of the GNU General Public License that is incorporated pursuant to the
+following paragraph.
+
+  Notwithstanding any other provision of this License, you have
+permission to link or combine any covered work with a work licensed
+under version 3 of the GNU General Public License into a single
+combined work, and to convey the resulting work.  The terms of this
+License will continue to apply to the part which is the covered work,
+but the work with which it is combined will remain governed by version
+3 of the GNU General Public License.
+
+  14. Revised Versions of this License.
+
+  The Free Software Foundation may publish revised and/or new versions of
+the GNU Affero General Public License from time to time.  Such new versions
+will be similar in spirit to the present version, but may differ in detail to
+address new problems or concerns.
+
+  Each version is given a distinguishing version number.  If the
+Program specifies that a certain numbered version of the GNU Affero General
+Public License "or any later version" applies to it, you have the
+option of following the terms and conditions either of that numbered
+version or of any later version published by the Free Software
+Foundation.  If the Program does not specify a version number of the
+GNU Affero General Public License, you may choose any version ever published
+by the Free Software Foundation.
+
+  If the Program specifies that a proxy can decide which future
+versions of the GNU Affero General Public License can be used, that proxy's
+public statement of acceptance of a version permanently authorizes you
+to choose that version for the Program.
+
+  Later license versions may give you additional or different
+permissions.  However, no additional obligations are imposed on any
+author or copyright holder as a result of your choosing to follow a
+later version.
+
+  15. Disclaimer of Warranty.
+
+  THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
+APPLICABLE LAW.  EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
+HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
+OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
+THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+PURPOSE.  THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
+IS WITH YOU.  SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
+ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
+
+  16. Limitation of Liability.
+
+  IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
+WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
+THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
+GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
+USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
+DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
+PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
+EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
+SUCH DAMAGES.
+
+  17. Interpretation of Sections 15 and 16.
+
+  If the disclaimer of warranty and limitation of liability provided
+above cannot be given local legal effect according to their terms,
+reviewing courts shall apply local law that most closely approximates
+an absolute waiver of all civil liability in connection with the
+Program, unless a warranty or assumption of liability accompanies a
+copy of the Program in return for a fee.
+
+                     END OF TERMS AND CONDITIONS
+
+            How to Apply These Terms to Your New Programs
+
+  If you develop a new program, and you want it to be of the greatest
+possible use to the public, the best way to achieve this is to make it
+free software which everyone can redistribute and change under these terms.
+
+  To do so, attach the following notices to the program.  It is safest
+to attach them to the start of each source file to most effectively
+state the exclusion of warranty; and each file should have at least
+the "copyright" line and a pointer to where the full notice is found.
+
+    <one line to give the program's name and a brief idea of what it does.>
+    Copyright (C) <year>  <name of author>
+
+    This program is free software: you can redistribute it and/or modify
+    it under the terms of the GNU Affero General Public License as published by
+    the Free Software Foundation, either version 3 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU Affero General Public License for more details.
+
+    You should have received a copy of the GNU Affero General Public License
+    along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+Also add information on how to contact you by electronic and paper mail.
+
+  If your software can interact with users remotely through a computer
+network, you should also make sure that it provides a way for users to
+get its source.  For example, if your program is a web application, its
+interface could display a "Source" link that leads users to an archive
+of the code.  There are many ways you could offer source, and different
+solutions will be better for different programs; see section 13 for the
+specific requirements.
+
+  You should also get your employer (if you work as a programmer) or school,
+if any, to sign a "copyright disclaimer" for the program, if necessary.
+For more information on this, and how to apply and follow the GNU AGPL, see
+<https://www.gnu.org/licenses/>.

README.md

@@ -1,145 +1,259 @@
-# OSTP (Ospab Stealth Transport Protocol)
+# OSTP - Ospab Stealth Transport Protocol
 
-OSTP is a high-throughput, robust, and multiplexed transport protocol engineered for secure, distributed industrial telemetry replication and real-time metric synchronization over unreliable, lossy networks. By implementing granular keystream scrambling and adaptive block framing, OSTP ensures absolute structural integrity and uniform entropy across all transmitted grid data, eliminating distinct traffic signatures and protecting assets against unauthorized analysis.
+[Русский язык](README.ru.md) · [Wiki](https://github.com/ospab/ostp/wiki) · [Contributing](CONTRIBUTING.md) · [Releases](https://github.com/ospab/ostp/releases)
+
+![GitHub Release](https://img.shields.io/github/v/release/ospab/ostp?style=for-the-badge&color=blue)
+![License: AGPL v3](https://img.shields.io/badge/License-AGPL%20v3-blue.svg?style=for-the-badge)
+![Platform: Windows | Linux | macOS | Android](https://img.shields.io/badge/Platform-Windows%20%7C%20Linux%20%7C%20macOS%20%7C%20Android-green.svg?style=for-the-badge)
+![Crypto](https://img.shields.io/badge/Crypto-Noise__NNpsk0-blueviolet?style=for-the-badge)
+![Transport](https://img.shields.io/badge/Transport-UDP%20ARQ-informational?style=for-the-badge)
+
+> A fast, custom encrypted transport protocol written in Rust.
+
+**OSTP** (Ospab Stealth Transport Protocol) is a high-performance transport protocol. It implements a custom ARQ transport over UDP, as well as a UoT (UDP-over-TCP) mode. Every byte on the wire - including packet headers - is cryptographically indistinguishable from random noise, making it highly resistant to Deep Packet Inspection (DPI).
 
 ---
 
-## Industrial Architecture
-
-The pipeline utilizes a highly optimized modular framework:
-- **ostp-core**: The foundational grid synchronization library hosting core transport primitives, keystream scrambling pipelines, Noise Protocol Framework cryptography, and zero-copy framed processing.
-- **ostp**: The consolidated cross-platform node daemon configured either as a telemetry collector (`server`) or relay bridge (`client`).
-- **ostp-jni**: Consolidated bindings allowing secure deployment of telemetry nodes across Android-embedded field equipment.
-
----
-
-## Feature Specification
-
-- **Keystream Scrambling (Entropy Masking)**: Internal packet fields are processed via high-entropy masking derived dynamically per session, ensuring absolute payload uniformity. This makes active traffic fully transparent to statistical network analyzers.
-- **Persistent Connection Multiplexing**: Enables high-fidelity continuous data channels, supporting parallel session structures and maintaining state persistence across volatile network interface rotations.
-- **Resilient Network Handoff**: Automatically detects and preserves active TCP pipelines when node endpoints experience topological shifts (e.g., cellular to fiber gateways) without interrupting upper-tier protocols.
-- **Pre-Shared Cryptographic Handshake**: Employs `Noise_NNpsk0_25519_ChaChaPoly_BLAKE2s` to validate remote nodes, establishing authentic channels instantly with post-quantum grade forward secrecy.
-- **Gateway Routing Protocol Support**: Standard dual-mode interfaces for legacy application routing via industrial SOCKS5/HTTP-CONNECT translation models.
-- **Static/Adaptive Block Shaping**: Eliminates behavioral data leaks through cryptographically randomized block-alignment schemes to maintain constant channel densities.
-
----
-
-## Provisioning and Configuration
-
-### Automated Linux Host Deployment
-
-For rapid, interactive provisioning on standard Linux host environments (x64/ARM), execute the unified installer via a single command:
+## Quick Install
 
+### Linux
 ```bash
 bash <(curl -Ls https://raw.githubusercontent.com/ospab/ostp/master/scripts/install.sh)
 ```
 
-*This routine autonomously fetches the correct stable release, registers a resilient systemd daemon, protects active user credentials during upgrades, and provides an interactive node setup.*
-
-### Automated Windows Host Deployment
-
-To perform quick installations, automated hot-swap updates, and set up global environment routes on Windows (x64/ARM64), run the following one-liner from an **elevated PowerShell console** (Run as Administrator):
-
+### Windows (PowerShell, run as Administrator)
 ```powershell
 irm https://raw.githubusercontent.com/ospab/ostp/master/scripts/install.ps1 | iex
 ```
 
-*This native script deploys the runtime directly to `C:\opt\ostp`, performs dynamic JSON template integrations, transparently handles active binary locks, and maps the node to the system environment PATH.*
+### Manual Download
+Download pre-built binaries for your platform from [GitHub Releases](https://github.com/ospab/ostp/releases).
 
-### Manual Node Initialization
+---
 
-The consolidated `ostp` daemon automates node certificate generation and base configuration templating.
+## Key Features
 
-**Provision Collector Node (Server):**
-```bash
-./ostp --init server
-```
-*This provisions `config.json` bound to an automated listening grid port with randomized secure node validation keys.*
+| Feature | Description |
+|---------|-------------|
+| **Full Traffic Obfuscation** | Every packet - including headers - is indistinguishable from random noise. Session IDs and nonces are masked with per-packet HMAC-derived keys. |
+| **Noise Protocol Handshake** | `Noise_NNpsk0_25519_ChaChaPoly_BLAKE2s` - PSK-authenticated, forward-secret key exchange with no static identity exposure. |
+| **Reliable UDP (ARQ)** | Selective ACK/NACK with rate-limited retransmission, configurable reorder buffer, and exponential backoff. |
+| **Multiplexed Streams** | Multiple logical TCP streams over a single encrypted UDP session with per-stream flow control. |
+| **Seamless Roaming** | Clients can switch networks (WiFi ↔ LTE) without session interruption - tracked by session-ID, not IP. |
+| **Management API** | Built-in REST API for third-party panels (3x-ui, custom dashboards). Per-user stats, traffic limits, key CRUD. |
+| **Fallback Server** | TCP fallback proxy to a web server - makes OSTP indistinguishable from nginx during active probing. |
+| **Multi-Listener** | Bind to multiple addresses simultaneously (dual-stack IPv4/IPv6, multi-port). |
+| **TUN Mode** | Full-system VPN via native `smoltcp` network stack without external dependencies. All traffic transparently routed through the tunnel. |
+| **UoT (UDP-over-TCP)** | Bare UDP-over-TCP tunnel, no protocol mimicry. Since all data is fully encrypted and length-prefixed, it bypasses DPI filters that block unknown UDP traffic by riding over a plain TCP connection. |
+| **Mobile & Web Apps** | Beautiful cross-platform mobile client (Flutter) and a modern Web Control Panel (React/Vite) for effortless server and client management. |
+| **TURN Relay** | RFC 5766 TURN support for environments where direct UDP is blocked. |
+| **Hot-Reload** | Runtime config reload without restart (access keys, exclusions, mux settings). |
+| **Structured Logging** | `tracing`-based logging with `RUST_LOG` filtering. JSON/file/syslog output support. |
+| **Cross-Platform** | Windows, Linux, macOS, Android, FreeBSD, MIPS, RISC-V. Single binary, no runtime dependencies. |
 
-**Provision Relay Node (Client):**
-```bash
-./ostp --init client
-```
+---
 
-### Node Integration Config
+## Architecture
 
-Configuration parameters are defined within `config.json` aligned adjacent to the service binary.
+```mermaid
+flowchart LR
+    %% Styles
+    classDef userApp fill:#e1f5fe,stroke:#01579b,stroke-width:2px,color:#01579b
+    classDef ostpCore fill:#e8f5e9,stroke:#2e7d32,stroke-width:2px,color:#2e7d32
+    classDef network fill:#fff3e0,stroke:#e65100,stroke-width:2px,color:#e65100,stroke-dasharray: 5 5
+    classDef external fill:#f3e5f5,stroke:#4a148c,stroke-width:2px,color:#4a148c
+    classDef fallback fill:#ffebee,stroke:#c62828,stroke-width:2px,color:#c62828
 
-#### Telemetry Collector Configuration (`config.json`)
-```json
-{
-  "mode": "server",
-  "listen": "0.0.0.0:50000",
-  "access_keys": [
-    "secure_node_registration_key_here"
-  ],
-  "debug": false
-}
-```
+    subgraph Local["💻 Client Device"]
+        Apps["Web Browser / Apps"]:::userApp
+        Socks["SOCKS5 / HTTP Proxy"]:::ostpCore
+        Tun["Global TUN (VPN)"]:::ostpCore
+        Client["OSTP Client Protocol Engine\n(Noise + ChaCha20 + ARQ)"]:::ostpCore
 
-#### Relay Bridge Configuration (`config.json`)
-```json
-{
-  "mode": "client",
-  "server": "COLLECTOR_ENDPOINT_IP:50000",
-  "access_key": "secure_node_registration_key_here",
-  "socks5_bind": "127.0.0.1:1088",
-  "tun": {
-    "enable": false,
-    "wintun_path": "./wintun.dll",
-    "ipv4_address": "10.1.0.2/24"
-  },
-  "exclude": {
-    "domains": [
-      "internal-system.lan",
-      "local.lan"
-    ],
-    "ips": [
-      "192.168.1.0/24",
-      "10.0.0.0/8"
-    ],
-    "processes": [
-      "local_monitoring.exe"
-    ]
-  },
-  "mux": {
-    "enabled": true,
-    "sessions": 2
-  }
-}
-```
+        Apps -->|TCP/UDP| Socks
+        Apps -->|IP Packets| Tun
+        Socks --> Client
+        Tun --> Client
+    end
 
-> [!IMPORTANT]
-> **TUN Mode & Privileges**
-> To route all global operating system traffic through OSTP, change `"enable": false` to `true` in the `tun` object.
-> - **Windows**: Requires running `ostp` as Administrator. The pre-packaged `tun2socks.exe` dependency must be located next to the `ostp` binary.
-> - **Linux**: Requires running `ostp` as root. The `tun2socks` and `iproute2` packages must be available in the system PATH or alongside the binary.
+    subgraph Internet["🌐 Hostile Network (DPI/Firewall)"]
+        Tunnel{"Fully Obfuscated\nEncrypted UDP\n(Looks like noise)"}:::network
+    end
 
-### Execution Parameters
+    subgraph Remote["🖥️ Remote VPS (Server)"]
+        Server["OSTP Server Protocol Engine\n(Authentication & Decryption)"]:::ostpCore
+        Relay["Connection Multiplexer"]:::ostpCore
+        Fallback["Fake Website\n(Nginx/Caddy)"]:::fallback
+        Target["Open Internet\n(YouTube, Google, etc)"]:::external
 
-Initiate telemetry processing by assigning the active configuration target:
+        Server -->|Decrypted Traffic| Relay
+        Server -->|Active Probe / Scanner| Fallback
+        Relay -->|Clear Traffic| Target
+    end
 
-```bash
-./ostp --config config.json
+    Client <==> Tunnel <==> Server
 ```
 
 ---
 
-## Operation & Reliability Metrics
+## Quick Start
 
-### Stream Multiplexing (Mux)
-> [!IMPORTANT]
-> **Parallel multiplexing is fully supported.**
-> The pipeline executes parallel handshake processes seamlessly, routing independent stream structures via separate cryptographic tunnels to maximize throughput.
+### 1. Generate config
 
-### Exclusion Engines (Bypass Modules)
-> [!NOTE]
-> Real-time exclusion engines are fully operational. Configured IP subnets, local domains, and internal processes correctly route traffic natively to prevent local loop latencies.
+```bash
+# On your VPS (server):
+./ostp init server
+
+# On your machine (client):
+./ostp init client
+```
+
+### 2. Edit config
+
+**Server** - set your access keys:
+```jsonc
+{
+  "mode": "server",
+  "listen": "0.0.0.0:50000",
+  "access_keys": ["YOUR_SECRET_KEY"],
+  "api": { "enabled": true, "bind": "127.0.0.1:9090", "token": "admin-token" },
+  "fallback": { "enabled": false, "listen": "0.0.0.0:443", "target": "127.0.0.1:8080" }
+}
+```
+
+**Client** - point to your server:
+```jsonc
+{
+  "mode": "client",
+  "server": "YOUR_SERVER_IP:50000",
+  "access_key": "YOUR_SECRET_KEY",
+  "socks5_bind": "127.0.0.1:1088",
+  "transport": { "mode": "udp" },
+  "tun": { "enable": false, "dns": "1.1.1.1" }
+}
+```
+
+### 3. Run
+
+```bash
+./ostp                         # Uses config.json in current directory
+./ostp --config /path/to.json  # Custom config path
+./ostp check                   # Validate config without running
+./ostp gk                      # Generate a new access key
+./ostp links                   # Print client share links
+```
+
+### 4. Connect via share link (one-liner)
+```bash
+./ostp connect "ostp://ACCESS_KEY@server.com:50000?..."
+```
+
+> [!WARNING]
+> Always wrap the `ostp://...` link in quotes (`"`) so your terminal doesn't misinterpret special characters like `&` or `?`.
+
+---
+
+## Management API
+
+Built-in REST API for building panels and dashboards.
+
+```bash
+# Server status
+curl -H "Authorization: Bearer mytoken" http://127.0.0.1:9090/api/server/status
+
+# List all users with traffic stats  
+curl -H "Authorization: Bearer mytoken" http://127.0.0.1:9090/api/users
+
+# Create a user with 10GB traffic limit
+curl -X POST -H "Authorization: Bearer mytoken" \
+  -H "Content-Type: application/json" \
+  -d '{"limit_bytes": 10737418240}' \
+  http://127.0.0.1:9090/api/users
+```
+
+Full API reference: [Management API](https://github.com/ospab/ostp/wiki/Management-API)
+
+---
+
+## CLI Reference
+
+```
+ostp [--config <PATH>] [COMMAND]
+
+Commands:
+  run                    Run the daemon using the config file (default when no command is given)
+  connect <URL>          Connect once using a share link: ostp://KEY@HOST:PORT
+  setup                  Interactive setup wizard
+  init <MODE>            Generate a template config (server/client/relay)
+  check                  Validate the configuration file and exit
+  gk                     Generate a secure access key (alias: generate-key)
+    --format <FMT>         Key format: hex, base64 (default: hex)
+    -n, --count <N>        Number of keys to generate (default: 1)
+  links                  Print client share links from the server config
+  import <URL>           Import a share link into the config file
+  update                 Update OSTP to the latest release
+    -b, --branch <NAME>    Release channel: stable, beta, alpha (default: stable)
+    -v, --version <VER>    Update to an exact version instead of the channel's latest
+  migrate                Force-migrate the configuration file to the current format
+  proxy-env              Print shell export commands for the local SOCKS proxy
+  proxy-env-clear        Print shell export commands to unset it
+  uninstall              Stop the service and remove the binary and config
+
+Global options:
+  --config <PATH>        Config file path (default: config.json)
+```
+
+Every subcommand also accepts `-h`/`--help` for its own option list.
+
+---
+
+## Protocol Summary
+
+| Layer | Mechanism |
+|-------|-----------|
+| Key Exchange | Noise NNpsk0 (X25519 + ChaChaPoly + BLAKE2s) zero-RTT |
+| Encryption | ChaCha20-Poly1305 AEAD per-packet |
+| Header Obfuscation | HMAC-SHA256 derived per-packet mask |
+| Reliability | Selective ACK with cumulative + SACK ranges |
+| Retransmission | Rate-limited NACK + exponential backoff RTO |
+| Keepalive | Ping/Pong with RTT measurement every 5s |
+
+---
+
+## Building from Source
+
+```bash
+# Prerequisites: Rust 1.75+
+cargo build --release
+
+# Cross-compile for Linux
+cross build --release --target x86_64-unknown-linux-gnu
+
+# Run tests
+cargo test -p ostp-core -p ostp-server
+```
+
+---
+
+## Documentation
+
+- **[Wiki](https://github.com/ospab/ostp/wiki)** - Full documentation
+- [Installation](https://github.com/ospab/ostp/wiki/Installation)
+- [Configuration Reference](https://github.com/ospab/ostp/wiki/Configuration)
+- [Management API](https://github.com/ospab/ostp/wiki/Management-API)
+- [Protocol Design](https://github.com/ospab/ostp/wiki/Protocol-Design)
+- [Building from Source](https://github.com/ospab/ostp/wiki/Building-from-Source)
+- [FAQ](https://github.com/ospab/ostp/wiki/FAQ)
 
 ---
 
 ## License
 
-OSTP is published under the Business Source License 1.1 (BSL), permitting unrestricted personal, non-commercial, and private utility deployments. This license automatically transitions to the permissive MIT License on May 14, 2030.
+GNU Affero General Public License v3.0 (AGPL-3.0). See [LICENSE](LICENSE) for the full text.
 
-For full licensing terms, refer to the accompanying [LICENSE](LICENSE) file or the official repository at [https://github.com/ospab/ostp](https://github.com/ospab/ostp).
+---
+
+## Contact
+
+- **Telegram**: [@ospab0](https://t.me/ospab0)
+- **Email**: gvoprgrg@gmail.com

README.ru.md

@@ -1,129 +1,245 @@
-# OSTP (Ospab Stealth Transport Protocol)
+# OSTP - Ospab Stealth Transport Protocol
 
-OSTP — это высокопроизводительный, надежный мультиплексируемый транспортный протокол, спроектированный для безопасной распределенной репликации промышленной телеметрии и синхронизации системных метрик реального времени в условиях нестабильных и зашумленных сетей передачи данных. За счет применения матричного маскирования сигнатурных потоков и адаптивного выравнивания границ блоков, OSTP гарантирует абсолютную структурную однородность и равномерную энтропию передаваемых данных, исключая появление статистических отпечатков трафика и защищая инфраструктуру от несанкционированного анализа.
+[English](README.md) · [Contributing](CONTRIBUTING.ru.md)
+
+![GitHub Release](https://img.shields.io/github/v/release/ospab/ostp?style=for-the-badge&color=blue)
+![License: AGPL v3](https://img.shields.io/badge/License-AGPL%20v3-blue.svg?style=for-the-badge)
+![Platform: Windows | Linux | macOS | Android](https://img.shields.io/badge/Platform-Windows%20%7C%20Linux%20%7C%20macOS%20%7C%20Android-green.svg?style=for-the-badge)
+![Crypto](https://img.shields.io/badge/Crypto-Noise__NNpsk0-blueviolet?style=for-the-badge)
+![Transport](https://img.shields.io/badge/Transport-UDP%20ARQ-informational?style=for-the-badge)
+
+> Быстрый кастомный зашифрованный транспортный протокол на Rust.
+
+**OSTP** (Ospab Stealth Transport Protocol) - кастомный транспортный протокол. Реализует собственный ARQ-транспорт поверх UDP, а также режим UoT (UDP-over-TCP). Каждый байт, включая заголовки пакетов, криптографически неотличим от случайного шума, что делает его устойчивым к системам глубокого анализа трафика (DPI).
 
 ---
 
-## Архитектура системы
+## Возможности
 
-Платформа построена на базе высокооптимизированного модульного каркаса:
-- **ostp-core**: Базовая библиотека синхронизации, обеспечивающая логику транспорта, алгоритмы маскирования энтропии, криптографическую обвязку на базе Noise Protocol Framework и потоковую обработку без копирования данных.
-- **ostp**: Унифицированный кроссплатформенный демон сетевого узла, конфигурируемый либо в режиме сборщика телеметрии (`server`), либо в режиме моста ретрансляции (`client`).
-- **ostp-jni**: Готовые связки для встраивания и развертывания сетевых узлов на базе оборудования под управлением ОС Android.
+| Возможность | Описание |
+|-------------|----------|
+| **Обфускация трафика** | Каждый пакет, включая заголовки, неотличим от случайного шума. Session ID и nonce маскируются HMAC-ключами, уникальными для каждого пакета. |
+| **Noise Protocol** | `Noise_NNpsk0_25519_ChaChaPoly_BLAKE2s` - аутентификация через PSK, forward secrecy, без раскрытия идентичности. |
+| **Reliable UDP (ARQ)** | Selective ACK/NACK с rate-limited ретрансмиссией, настраиваемым reorder-буфером и exponential backoff. Разработан для 10 Гбит/с. |
+| **Мультиплексирование** | Несколько логических TCP-потоков поверх одной зашифрованной UDP-сессии с per-stream flow control. |
+| **Бесшовный роуминг** | Клиент может менять сети (WiFi ↔ 4G) без разрыва сессии - сервер отслеживает session-ID, а не IP-адрес. |
+| **TUN-режим** | Полносистемный VPN без внешних зависимостей (встроенный network stack на базе `smoltcp`). |
+| **UoT (UDP-over-TCP)** | Голый туннель UDP-over-TCP, без имитации протоколов. Поскольку все данные полностью зашифрованы и имеют префикс длины, он обходит DPI фильтры, блокирующие неизвестный UDP трафик, передавая всё по обычному TCP соединению. |
+| **Мобильные и Web приложения** | Красивый кроссплатформенный мобильный клиент (Flutter) и современная Web панель управления (React/Vite) для удобного администрирования. |
+| **TURN Relay** | RFC 5766 TURN для окружений, где прямой UDP заблокирован. |
+| **Hot-Reload** | Перезагрузка конфига в рантайме без перезапуска (ключи, исключения, mux, TURN). |
+| **Кросс-платформа** | Windows, Linux, macOS, Android. Один бинарник, без зависимостей. |
 
 ---
 
-## Технические спецификации
+## Архитектура
 
-- **Маскирование энтропии (Скрытие сигнатур)**: Внутренние поля пакетов проходят динамическую высокоэнтропийную потоковую обработку на каждом сеансе связи, обеспечивая предельную однородность трафика. Это делает сетевые потоки невидимыми для автоматических анализаторов топологии.
-- **Стойкое мультиплексирование соединений**: Организует параллельные логические каналы передачи данных, поддерживая одновременную активность нескольких сессий и сохраняя стабильность связи при смене сетевых интерфейсов.
-- **Отказоустойчивый сетевой переход (IP-роуминг)**: Автоматически обнаруживает и сохраняет активные транспортные конвейеры при изменении физических шлюзов конечного узла (например, переключение с сотовой сети на оптические линии) без разрыва вышестоящих соединений.
-- **Безопасное рукопожатие (PSK Handshake)**: Использует схему `Noise_NNpsk0_25519_ChaChaPoly_BLAKE2s` для аутентификации удаленных узлов, обеспечивая мгновенный запуск защищенного канала с гарантиями совершенной прямой секретности (Forward Secrecy).
-- **Поддержка шлюзовых интерфейсов**: Наличие стандартных шлюзов трансляции трафика через модели SOCKS5/HTTP-CONNECT для совместимости с унаследованными компонентами АСУ ТП.
-- **Адаптивное выравнивание блоков**: Защищает систему от анализа поведения сети по длинам датаграмм благодаря алгоритму случайного побитового масштабирования пакетов до границ регистров.
+```mermaid
+flowchart LR
+    %% Styles
+    classDef userApp fill:#e1f5fe,stroke:#01579b,stroke-width:2px,color:#01579b
+    classDef ostpCore fill:#e8f5e9,stroke:#2e7d32,stroke-width:2px,color:#2e7d32
+    classDef network fill:#fff3e0,stroke:#e65100,stroke-width:2px,color:#e65100,stroke-dasharray: 5 5
+    classDef external fill:#f3e5f5,stroke:#4a148c,stroke-width:2px,color:#4a148c
+    classDef fallback fill:#ffebee,stroke:#c62828,stroke-width:2px,color:#c62828
+
+    subgraph Local["💻 Устройство клиента"]
+        Apps["Браузер / Приложения"]:::userApp
+        Socks["SOCKS5 / HTTP Прокси"]:::ostpCore
+        Tun["Global TUN (VPN)"]:::ostpCore
+        Client["OSTP Клиент\n(Noise + ChaCha20 + ARQ)"]:::ostpCore
+
+        Apps -->|TCP/UDP| Socks
+        Apps -->|IP Пакеты| Tun
+        Socks --> Client
+        Tun --> Client
+    end
+
+    subgraph Internet["🌐 Сеть с цензурой (DPI)"]
+        Tunnel{"Зашифрованный UDP\n(Выглядит как белый шум)"}:::network
+    end
+
+    subgraph Remote["🖥️ Удаленный сервер (VPS)"]
+        Server["OSTP Сервер\n(Аутентификация)"]:::ostpCore
+        Relay["Мультиплексор соединений"]:::ostpCore
+        Fallback["Фейковый сайт\n(Nginx/Caddy)"]:::fallback
+        Target["Свободный интернет\n(YouTube, Google и т.д.)"]:::external
+
+        Server -->|Расшифрованный трафик| Relay
+        Server -->|Сканеры цензоров| Fallback
+        Relay -->|Чистый трафик| Target
+    end
+
+    Client <==> Tunnel <==> Server
+```
 
 ---
 
-## Развертывание и настройка
-
-### Автоматическая установка на Linux (Рекомендуется)
-
-Для быстрого интерактивного развертывания узла в стандартных серверных средах Linux выполните команду установки непосредственно в консоли терминала:
+## Установка
 
+### Linux
 ```bash
 bash <(curl -Ls https://raw.githubusercontent.com/ospab/ostp/master/scripts/install.sh)
 ```
 
-*Данный сценарий автономно загружает подходящий бинарный релиз, регистрирует системную службу демона в операционной системе и интерактивно настраивает конфигурационные шаблоны с помощью встроенных инструментов компиляции.*
-
-### Ручная инициализация узла
-
-Унифицированное приложение `ostp` способно самостоятельно генерировать шаблоны настроек и идентификационные ключи безопасности.
-
-**Инициализация узла сборщика (Сервер):**
-```bash
-./ostp --init server
-```
-*Эта команда создает файл `config.json`, привязанный к автоматическому порту прослушивания, и записывает туда сгенерированные случайные ключи авторизации.*
-
-**Инициализация узла моста (Клиент):**
-```bash
-./ostp --init client
-```
-
-### Конфигурация интеграции
-
-Рабочие параметры узла задаются в файле `config.json`, расположенном рядом с исполняемым файлом демона.
-
-#### Конфигурация сборщика телеметрии (`config.json`)
-```json
-{
-  "mode": "server",
-  "listen": "0.0.0.0:50000",
-  "access_keys": [
-    "secure_node_registration_key_here"
-  ],
-  "debug": false
-}
-```
-
-#### Конфигурация моста ретрансляции (`config.json`)
-```json
-{
-  "mode": "client",
-  "server": "COLLECTOR_ENDPOINT_IP:50000",
-  "access_key": "secure_node_registration_key_here",
-  "socks5_bind": "127.0.0.1:1088",
-  "tun": {
-    "enable": false,
-    "wintun_path": "./wintun.dll",
-    "ipv4_address": "10.1.0.2/24"
-  },
-  "exclude": {
-    "domains": [
-      "internal-system.lan",
-      "local.lan"
-    ],
-    "ips": [
-      "192.168.1.0/24",
-      "10.0.0.0/8"
-    ],
-    "processes": [
-      "local_monitoring.exe"
-    ]
-  },
-  "mux": {
-    "enabled": true,
-    "sessions": 2
-  }
-}
-```
-
-### Запуск узла
-
-Для активации процессов обмена телеметрией запустите приложение с указанием пути к активному файлу параметров:
-
-```bash
-./ostp --config config.json
+### Windows (PowerShell от Администратора)
+```powershell
+irm https://raw.githubusercontent.com/ospab/ostp/master/scripts/install.ps1 | iex
 ```
 
 ---
 
-## Метрики стабильности и производительности
+## Конфигурация
 
-### Мультиплексирование потоков (Mux)
-> [!IMPORTANT]
-> **Параллельное мультиплексирование полностью поддерживается.**
-> Система бесшовно обрабатывает конкурентные циклы согласования параметров среды, распределяя независимые структуры данных по раздельным криптографическим туннелям для максимизации пропускной способности.
+Создать конфиг по умолчанию:
+```bash
+./ostp init server   # VPS
+./ostp init client   # Локальная машина
+```
 
-### Модули исключений (Bypass Engines)
-> [!NOTE]
-> Механизмы маршрутизации в обход шины передачи полностью готовы к эксплуатации. Указанные в конфигурации IP-подсети, локальные доменные зоны и процессы корректно направляются напрямую через штатный сетевой стек ОС, исключая дополнительные задержки маршрутов.
+### Сервер (`config.json`)
+```jsonc
+{
+  "mode": "server",
+  "listen": "0.0.0.0:50000",
+  "access_keys": ["ВАШ_КЛЮЧ"],
+  "debug": false,
+  // Опционально: проксировать трафик через upstream
+  "outbound": {
+    "enabled": false,
+    "protocol": "socks5",
+    "address": "127.0.0.1",
+    "port": 9050,
+    "default_action": "proxy"
+  }
+}
+```
+
+### Клиент (`config.json`)
+```jsonc
+{
+  "mode": "client",
+  "server": "IP_СЕРВЕРА:50000",
+  "access_key": "ВАШ_КЛЮЧ",
+  "socks5_bind": "127.0.0.1:1088",
+  "debug": false,
+  // Настройки транспорта (udp или uot)
+  "transport": {
+    "mode": "udp"
+  },
+  // TUN-режим (полносистемный VPN)
+  "tun": {
+    "enable": false,
+    "dns": "1.1.1.1"
+  },
+  // Мультиплексирование: несколько UDP-сессий
+  "mux": {
+    "enabled": false,
+    "sessions": 2
+  },
+  // TURN-реле для заблокированных сетей
+  "turn": {
+    "enabled": false,
+    "server_addr": "turn.example.com:3478",
+    "username": "user",
+    "access_key": "pass"
+  },
+  // Исключения (идут напрямую, минуя туннель)
+  "exclude": {
+    "domains": ["example.local"],
+    "ips": ["192.168.0.0/16"]
+  }
+}
+```
+
+---
+
+## Использование
+
+```bash
+# Запуск с конфигом
+./ostp --config config.json
+
+# Или просто (ищет config.json рядом с бинарником)
+./ostp
+```
+
+### Справка по командам
+
+```
+ostp [--config <PATH>] [КОМАНДА]
+
+Команды:
+  run                    Запустить демон по конфигу (по умолчанию, если команда не указана)
+  connect <URL>          Подключиться по share-ссылке: ostp://KEY@HOST:PORT
+  setup                  Интерактивный мастер настройки
+  init <MODE>            Сгенерировать шаблон конфига (server/client/relay)
+  check                  Проверить конфиг и выйти
+  gk                     Сгенерировать access-key (алиас: generate-key)
+    --format <FMT>         Формат ключа: hex, base64 (по умолчанию hex)
+    -n, --count <N>        Количество ключей (по умолчанию 1)
+  links                  Вывести client-share-ссылки из серверного конфига
+  import <URL>           Импортировать share-ссылку в конфиг
+  update                 Обновить OSTP до актуального релиза
+    -b, --branch <NAME>    Канал релиза: stable, beta, alpha (по умолчанию stable)
+    -v, --version <VER>    Обновиться на точную версию вместо последней в канале
+  migrate                Принудительно мигрировать конфиг к текущему формату
+  proxy-env              Вывести shell-команды для локального SOCKS-прокси
+  proxy-env-clear        Вывести shell-команды для их отмены
+  uninstall              Остановить сервис и удалить бинарник с конфигом
+
+Глобальные опции:
+  --config <PATH>        Путь к конфигу (по умолчанию config.json)
+```
+
+У каждой подкоманды есть своя справка через `-h`/`--help`.
+
+### TUN-режим (Windows)
+Использует встроенный сетевой стек `smoltcp` и виртуальный адаптер `wintun` (необходима `wintun.dll`). Требует запуска с правами Администратора.
+
+### TUN-режим (Linux)
+Использует встроенный сетевой стек `smoltcp` и `/dev/net/tun`. Требует запуска от имени `root` (или наличия `CAP_NET_ADMIN`).
+
+---
+
+## Спецификация протокола
+
+| Уровень | Механизм |
+|---------|----------|
+| Обмен ключами | Noise NNpsk0 (X25519 + ChaChaPoly + BLAKE2s) zero-RTT |
+| Шифрование | ChaCha20-Poly1305 AEAD на каждый пакет |
+| Обфускация заголовков | HMAC-SHA256 маска session_id + nonce, уникальная для каждого пакета |
+| Надёжность | Selective ACK с cumulative + SACK диапазонами |
+| Ретрансмиссия | Rate-limited NACK (30мс cooldown) + exponential backoff RTO |
+| Flow Control | Окно in-flight (только retransmittable фреймы) |
+| Keepalive | Ping/Pong с измерением RTT каждые 5с |
+| Таймаут сессии | 60с на клиенте, 300с на сервере |
+
+---
+
+## Сборка из исходников
+
+```bash
+# Требования: Rust toolchain (1.75+)
+cargo build --release
+
+# Кросс-компиляция для Linux
+cross build --release --target x86_64-unknown-linux-gnu
+```
+
+---
+
+## Документация
+
+- [Архитектура](docs/ru/architecture.md)
+- [Спецификация протокола](docs/ru/specification.md)
+- [Дизайн обфускации](docs/ru/obfuscation.md)
+- [Администрирование сервера](docs/ru/server.md)
+- [Настройка клиента](docs/ru/client.md)
+- [Интеграции](docs/ru/integrations.md)
 
 ---
 
 ## Лицензия
 
-OSTP публикуется на условиях лицензии Business Source License 1.1 (BSL), которая разрешает неограниченное личное, некоммерческое и частное использование протокола. С 14 мая 2030 года лицензия автоматически переходит в категорию открытого ПО с разрешительной лицензией MIT.
-
-С полным текстом лицензионного соглашения можно ознакомиться в приложенном файле [LICENSE](LICENSE) или в официальном репозитории проекта по адресу [https://github.com/ospab/ostp](https://github.com/ospab/ostp).
+GNU Affero General Public License v3.0 (AGPL-3.0). Полный текст - в файле [LICENSE](LICENSE).

REBUILD_PLAN.md

@@ -0,0 +1,185 @@
+# Чистая переборка на базе v0.2.98
+
+База: `v0.2.98` (commit `31d0020`) — последняя версия, которая **стабильно работает**.
+Ветка: `clean-rebuild`. Всё, что появилось после (0.3.1 … 0.3.21), переносим
+**выборочно и с чистой головой**, а не копируем рефактор целиком.
+
+Принцип: 0.3.1 принёс «модульный multi-server рефактор» + лавину фич — и вместе с
+ними нестабильность. Берём только проверенное и нужное.
+
+---
+
+## Решения (зафиксировано пользователем)
+- **Junk-пакеты + TCP-фрагментация — ОСТАВЛЯЕМ** (нравятся). НО починить вредную
+  часть: junk по UDP не должен выглядеть для сервера как `Unauthorized probe`
+  (rate-limit/гейт на сервере), иначе флуд лога и риск самобана клиента. Фича
+  остаётся — чиним поведение, а не выпиливаем. Тонкая настройка — §E.
+- **Версия переборки — 0.4.0** (решено; 0.3.x сожжены в pre-release).
+- **WSS и Reality (TLS-мимикрия) — ВЫКИНУТЬ.** Путь проекта — **zapret-like**:
+  обфускация/DPI-evasion на уровне пакетов (junk, фрагментация, обфускация), а НЕ
+  мимикрия под TLS. Reality с нуля тяжела и не вписывается.
+- **Multi-server — НЕ НУЖЕН.** Режем до одного сервера → уходит urltest-группа и
+  половина сложности 0.3.1.
+- **Конфиг — ПЛОСКИЙ по сути, но оформлен красиво/секционно как сейчас** (решено).
+  Сохраняем читаемую секционную структуру (server / transport / tun / dns / exclude
+  и т.п.), но **выпиливаем модульную машинерию**: массивы `inbounds[]`/`outbounds[]`,
+  `routing.rules[]` с тегами, `default_outbound`, urltest, мульти-сервер. Один сервер
+  на конфиг. Исключения = плоский список внутри секции `exclude`.
+- **Профили — ОСТАВЛЯЕМ, single-select, в UI-слое** (решено). Профиль = сохранённый
+  конфиг одного сервера; активен ровно один (radio). Список/выбор/share живут во
+  фронте (prefs GUI / Flutter); **ядро о профилях не знает** — на «Подключить» из
+  выбранного профиля генерится плоский конфиг на один сервер. Никаких чекбоксов/
+  мульти-актив/urltest.
+- **Derived-secrets — ОСТАВЛЯЕМ, но ОБЯЗАТЕЛЬНО проверить, что он РЕАЛЬНО работает:**
+  старый клиент НЕ должен подключаться к новому серверу. В прошлой реализации это
+  НЕ соблюдалось (старый клиент → новый сервер подключался) — значит сервер всё ещё
+  принимал старый формат handshake / obfuscation-key. Это **баг**, закрыть в первую
+  очередь: сервер обязан отвергать всё, что не прошло derived-secrets.
+- **Лицензия — AGPLv3.**
+- **Брендинг — ОСТАВЛЯЕМ**: тёмная тема + орёл на фоне (watermark/логотип).
+- **Стелс-философия (north-star): zapret-like** — «нет узнаваемого заголовка +
+  манипуляции пакетами» (обфускация, junk, фрагментация, DNS/UoT-транспорты), а НЕ
+  «притворись известным протоколом» (Reality/WSS — выкинуты).
+
+---
+
+## 0. Корневая причина нестабильности 0.3.x
+**Модульный multi-server рефактор (0.3.1)** — `580faf6`, `8ed66f9`, `67f9c06`.
+Сменил формат конфига (inbounds/outbounds/routing/urltest), session-модель,
+hot-reload. Источник большинства багов (мёртвые маршруты, фейк-коннект,
+рассинхрон конфига). **НЕ копировать целиком.** Если multi-server реально нужен —
+добавлять минимально и поверх рабочей одно-серверной модели 0.2.98.
+
+---
+
+## A. ВЫКИНУТЬ / не переносить
+1. **WSS-фрейминг и Reality (TLS-мимикрия)** — оба выкинуть. Путь zapret-like, а не
+   маскировка под TLS-сайт; Reality (`reality.rs`) к тому же сложно сделать корректно
+   с нуля. Удалить из базы 0.2.98 целиком.
+2. **Multi-server / urltest-группа** — не нужен. Один сервер на конфиг.
+3. Остатки **tun2socks** на Android (`libtun2socks.so`, `tun2socks-arm64`,
+   `tun_child`, `t2sBinPath`) — давно мёртвый код, только раздувает APK. Не тащить.
+
+> ⚠️ Junk-пакеты и TCP-фрагментация **ОСТАЮТСЯ** (см. Решения и §E) — это уже не
+> «мусор». Но junk по UDP нужно сделать так, чтобы сервер его не считал
+> `Unauthorized probe` (rate-limit/гейт), иначе лог-флуд и риск самобана.
+
+---
+
+## B. ОБЯЗАТЕЛЬНО перенести (фиксы стабильности)
+- **fd limits / EMFILE** — `922cf0b`.
+- **Lifecycle хелпера**: принудительный `std::process::exit` после остановки, чтобы
+  не оставался зомби-процесс, держащий адаптер `ostp_tun` и дефолтный маршрут — `b6e78c1`.
+- **Bypass-маршрут сервера через `route.exe` по шлюзу** (а не legacy
+  `CreateIpForwardEntry`, который падал с err 160 из-за рассинхрона индексов
+  интерфейсов) — `b6e78c1`.
+- **IPC хелпера** (ChaCha20Poly1305 + hex) + **единый формат логов** — `ee38b15`.
+- **Closing-state fix** + `sent_history` на `BTreeMap` (O(log n) NACK) — `47d44fa`.
+- **Handshake timeout fixes** — `d65af35`, `6eb7b36` (ждать ответ до отправки данных).
+- **Buffer / UDP handler** — `b5e830a`.
+- **Логи**: UoT и unauthorized-probe → debug; rate-limit probe-лога — `1151726`, `fc339b3`.
+
+---
+
+## C. Протокол / крипто — решить и перенести
+- **Derived secrets handshake** — `f8f27d3`. PSK и obfuscation-key выводятся из
+  access-key через HKDF; handshake-payload = `[timestamp][session_id][access_key]`;
+  параметры паддинга деривируются; timestamp anti-replay (±300с).
+  ⚠️ **Ломает совместимость с 0.2.98 wire** (старый клиент не подключится).
+  Безопаснее старого (raw-PSK + нулевой obfuscation-key). **РЕШЕНИЕ:** переносим ли
+  (тогда нужен ребилд всех клиентов) — ДА, скорее всего, но осознанно.
+- **l4_protocol** для server outbound — `2997bfd`, `ad3a8cb`, `aae9d22`.
+
+---
+
+## D. Транспорты — перенести аккуратно (большие куски)
+- **DNS transport (dnstt)** как fallback — `3f1adbc`, `3ced4a1`, `d031b15`,
+  `10c1772`, `b31da29`. Полезно против блокировок, но объёмно и со своей
+  фрагментацией/reassembly. Переносить отдельным изолированным модулем.
+- UoT (UDP-over-TCP) — уже есть в 0.2.98, проверить что не сломан.
+
+---
+
+## E. Тонкая настройка junk/фрагментации (как в AmneziaWG)
+Junk и фрагментацию **оставляем** (Решения), а это — их параметризация. Главное
+условие: **координация клиент↔сервер**, иначе junk превращается в probe-флуд.
+- `Jc` — кол-во junk-пакетов, `Jmin`/`Jmax` — размеры; **сервер знает и молча отбрасывает**.
+- `S1`/`S2` — размеры init/response подгоняются.
+- Магические заголовки/сигнатуры пакетов (`H1..H4`).
+Реализовать как явные настраиваемые поля (не хардкод). Сервер ОБЯЗАН их понимать.
+Сам факт junk/frag — в базе; это «желание» — сделать их настраиваемыми. Можно потом.
+
+---
+
+## F. GUI (desktop) — перенести нужное, без хаоса
+- Профили на странице **настроек** (пусто + «Create a new profile» + «+» когда нет
+  профиля; «+» → меню «из ссылки / вручную»). Главный экран не усложнять.
+- **Share** профиля: QR (генерить локально, ключ наружу не отдавать — крейт `qrcode`)
+  + копируемая `ostp://` ссылка.
+- **Метрики**: байты считать в TUN-инбаунде; rtt брать из round-trip handshake
+  (а не отдельным TCP-probe).
+- **Health/состояние**: «connected» по реальной достижимости сервера на ПРАВИЛЬНОМ
+  порту (не хардкод :443), а не по факту «процесс запустился».
+- **routing**: всегда задавать `default_outbound: "proxy"`; ключи правил —
+  `domain_suffix` / `ip_cidr` / `process_name` (не `domains/ips/processes`).
+- Смена сервера = полный **stop+start**, а не hot-reload (иначе остаётся старый сервер).
+- Никаких непрогарженных `addEventListener` на удалённые элементы (краш init).
+
+---
+
+## G. Мобилка (Flutter + JNI) — перенести нужное
+- **routing**: тот же `default_outbound` + правильные ключи правил
+  (без них трафик шёл мимо туннеля — реальный IP).
+- **Байты на Android**: считать в обеих задачах fd-пути (read=upload, write=download).
+- **rtt**: из handshake (health-probe сокет на Android не protected → до сервера не доходит).
+- **Смена сети (WiFi↔LTE)**: реальный reconnect (сейчас `notifyNetworkChanged` — no-op).
+- **fd ownership**: НЕ двойное закрытие (Rust `OwnedFd` + Kotlin `close()`) → `detachFd()`.
+- **Share** профиля: QR (`qr_flutter`) + ссылка.
+- Выкинуть tun2socks (см. §A.3).
+
+---
+
+## H. Инфра / лицензия / брендинг
+- Лицензия: **AGPLv3** ✅ (зафиксировано). В 0.2.98 был BSL 1.1 → заменить (`9ce9e6d`).
+- **Брендинг — ОСТАВЛЯЕМ** ✅: тёмная тема + орёл на фоне (watermark/логотип) в GUI.
+  Перенести из текущего `ostp-gui` (assets/logo.svg, тёмная палитра) в чистую переборку.
+- Панель/license-check: open-source без license-check — `5782107`, `99ff76d` (если нужно).
+- Версионирование/CI build-script — `774d926` и пр.
+
+---
+
+## Инвентаризация базы 0.2.98 (что уже есть / что портировать)
+- **Есть в 0.2.98**: WSS (→ удалить), Reality/`reality.rs` (→ удалить),
+  инфра derived-secrets (`derive_all_secrets`, `obfuscation_key`) — но клиент юзал
+  dummy-ключи до `f8f27d3`.
+- **Нет в 0.2.98 — портировать из пост-0.2.98 кода**: junk-пакеты, TCP-фрагментация,
+  DNS-transport (dnstt), фикс derived-secrets `f8f27d3`, все фиксы §B, GUI/мобилка §F/§G.
+
+## Что легко упустить (решить до старта)
+1. **Версия переборки — 0.4.0** (решено). Сожжённые 0.3.x не переиспользуем.
+2. **Серверный конфиг — тоже плоский** и согласован с клиентским. Сервер обязан
+   поддерживать всё оставленное: derived-secrets (и **отвергать** старый формат),
+   корректную обработку junk (не probe-флуд), UoT, DNS-transport, management API.
+3. **Версия/магический байт протокола ДО крипто-слоя.** Сейчас нельзя отличить старый
+   handshake от нового — отсюда баг «старый клиент → новый сервер подключился».
+   Добавить версию в wire → будущие изменения управляемы, сервер чётко режет
+   несовместимое. Это системный фикс проблемы derived-secrets.
+4. **Клиент и сервер обновляются ВМЕСТЕ** — derived-secrets ломает совместимость,
+   смешивать старое и новое нельзя. Координировать выкладку.
+5. **Reality — выкинуть** (решено; в базе 0.2.98 есть `reality.rs` → удалить целиком).
+6. **Verify-loop = критерий «готово».** Каждая фича проверяется реальным тестом, не
+   «на словах»: connect → `curl` показывает IP **сервера**; старый клиент к новому
+   серверу **не** подключается; смена сети на мобилке восстанавливает туннель.
+
+## Порядок переборки (предложение, 1 сессия)
+1. §B (фиксы стабильности) — на чистый 0.2.98.
+2. §C (derived-secrets) — и СРАЗУ проверить: старый клиент к новому серверу НЕ
+   подключается (в прошлый раз был баг — подключался).
+3. **Junk + TCP-фрагментация** — перенести (оставляем), но junk по UDP не должен
+   читаться сервером как `Unauthorized probe` (rate-limit/гейт на сервере).
+4. §F/§G по минимуму (routing, метрики, состояние, share) **+ брендинг** (тёмная
+   тема, орёл на фоне).
+5. §D (DNS transport) — если нужно.
+6. ВЫКИНУТЬ: **WSS, multi-server, tun2socks** (§A). §E (тюнинг junk) — позже.
+7. Конфиг: плоский по сути, секционно-оформленный, один сервер (РЕШЕНО — без
+   inbounds/outbounds/routing-движка).

app-icon.svg

@@ -0,0 +1,15 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 512 512">
+  <defs>
+    <linearGradient id="g2" x1="0%" y1="0%" x2="100%" y2="100%">
+      <stop offset="0%" stop-color="#111827" />
+      <stop offset="100%" stop-color="#374151" />
+    </linearGradient>
+    <linearGradient id="g2_path" x1="0%" y1="0%" x2="100%" y2="100%">
+      <stop offset="0%" stop-color="#3B82F6" />
+      <stop offset="100%" stop-color="#14B8A6" />
+    </linearGradient>
+  </defs>
+  <rect width="512" height="512" rx="120" fill="url(#g2)" />
+  <path d="M144 256c0-61.9 50.1-112 112-112s112 50.1 112 112-50.1 112-112 112S144 317.9 144 256zm-48 0c0 88.4 71.6 160 160 160s160-71.6 160-160S344.4 96 256 96 96 167.6 96 256z" fill="url(#g2_path)"/>
+  <circle cx="256" cy="256" r="40" fill="#F59E0B" />
+</svg>

docs/banner.txt

@@ -0,0 +1,7 @@
+
+  ____   _____ _______ _____  
+ / __ \ / ____|__   __|  __ \ 
+| |  | | (___    | |  | |__) |
+| |  | |\___ \   | |  |  ___/ 
+| |__| |____) |  | |  | |     
+ \____/|_____/   |_|  |_|     

docs/en/client.md

@@ -59,7 +59,7 @@ To minimize latency and overhead for trusted resources, the OSTP client incorpor
 
 ---
 
-## Multiplexing & Known Session Constraints
+## Multiplexing
 
 The wire protocol provides support for bundling multiple physical UDP session handles into a single logical transport pipeline via the `"mux"` block:
 
@@ -70,12 +70,5 @@ The wire protocol provides support for bundling multiple physical UDP session ha
 }
 ```
 
-### Current Implementation Limits:
-> [!WARNING]
-> **Currently, utilizing more than 1 multiplexed session (`sessions > 1`) is NOT supported and will result in complete traffic loss.**
->
-> **Observed Bug Behavior:**
-> If multiple sessions are initiated (e.g., `sessions: 3`), the client executes successful handshakes for each endpoint (yielding repeated `Connected UDP directly to` lines in diagnostic logs), and the server initializes the matching tracking slots. However, during the payload demultiplexing phase, the server pipeline fails to bridge payloads back to active streams, dropping all encapsulated packets.
-> 
-> **Resolution Requirement:**
-> You MUST ensure that the `"mux"` block remains disabled (`"enabled": false`) OR is manually constrained to exactly **1** session (`"sessions": 1`).
+### Current Status
+Multi-session multiplexing (`sessions > 1`) is supported. Use the `"mux"` block to scale concurrent transport sessions as needed for throughput or resiliency.

docs/en/obfuscation.md

@@ -5,40 +5,38 @@ Traditional tunneling protocols (such as TLS, OpenVPN, and WireGuard) exhibit di
 
 ---
 
-## Obfuscation Key Derivation
+## Secret Derivation
 
-To dynamically mask protocol data, an 8-byte obfuscation key is statically derived from the shared `access_key` configured on both the client and the server:
+Every protocol secret — the obfuscation key, the Noise PSK, the handshake padding range, and the per-key junk marker (see below) — is derived from the shared `access_key` via a single HKDF-SHA256 pass, domain-separated by a trailing info byte per output:
 
-$$\text{Key} = \text{SHA-256}(\text{access\_key})[0..8]$$
+```
+PRK              = HKDF-Extract(salt = SHA-256(access_key)[0..16], IKM = access_key || PROTOCOL_VERSION)
+obfuscation_key  = HKDF-Expand(PRK, info = SHA-256(access_key)[16..] || 0x01, 8 bytes)
+psk              = HKDF-Expand(PRK, info = SHA-256(access_key)[16..] || 0x02, 32 bytes)
+handshake_pad    = HKDF-Expand(PRK, info = SHA-256(access_key)[16..] || 0x03, 2 bytes)
+junk_marker      = HKDF-Expand(PRK, info = SHA-256(access_key)[16..] || 0x04, 4 bytes)
+```
 
-This key is established pre-session and is never transmitted across the wire in any capacity.
+The wire protocol version is mixed into the IKM, not sent as a plaintext byte: peers on a different protocol version derive an entirely different `obfuscation_key`, so they simply cannot deobfuscate each other's packets and are rejected as unauthorized — a hard version gate with no recognizable marker ever appearing on the wire. No secret is ever transmitted; both sides derive the same values independently from the shared access key.
 
 ---
 
 ## Dynamic In-Place Masking Algorithm
 
-OSTP datagrams are processed "in-place" immediately prior to transmission and right after arrival. Two distinct mathematical modes are utilized based on the current handshake phase:
+OSTP datagrams are masked "in-place" immediately prior to transmission and right after arrival. The mask itself is **derived from the packet's own ciphertext**, not from a fixed keystream or a counter, so it changes with every packet automatically:
+
+```
+mask = HMAC-SHA256(key = obfuscation_key, message = ciphertext[0..min(32, len)])
+```
 
 ### 1. Handshake Phase Mode (`is_handshake = true`)
-During connection initiation (Noise Handshake), the wire packet consists of a 4-byte `session_id` prefixed to the Noise payload. To mask the fixed session ID:
-
-*   **Masking**: The first 4 bytes are XORed with the first 4 bytes of the derived obfuscation key:
-    $$\text{raw}[i] = \text{raw}[i] \oplus \text{Key}[i \pmod 8], \quad i \in [0..3]$$
-*   **De-masking**: A repeated XOR with the identical key bytes recovers the original `session_id`.
+The wire packet is `[4-byte session_id][2-byte noise_len][Noise payload]`. The mask is computed over the Noise payload (`raw[6..]`), and its first 6 bytes are XORed onto `session_id || noise_len`.
 
 ### 2. Data Transmission Mode (`is_handshake = false`)
-Post-handshake, the wire layout contains:
-`[4-byte session_id]` + `[8-byte nonce]` + `[AEAD Ciphertext]`
+The wire packet is `[4-byte session_id][8-byte nonce][AEAD ciphertext]`. The mask is computed over the AEAD ciphertext, and its first 12 bytes are XORed onto `session_id || nonce`.
 
-To completely randomize metadata, a two-tiered dynamic XOR masking process is applied:
-
-1.  **Nonce Masking**: The 8-byte `nonce` (sequence counter) is XORed with the full 8-byte static key:
-    $$\text{nonce\_bytes}[i] = \text{nonce\_bytes}[i] \oplus \text{Key}[i], \quad i \in [0..7]$$
-2.  **Session ID Masking**: The 4-byte `session_id` is masked using high dynamic entropy — the lower 32 bits of the **original (unmasked)** `nonce` value:
-    $$\text{session\_id\_bytes}[i] = \text{session\_id\_bytes}[i] \oplus \text{real\_nonce\_low32\_bytes}[i], \quad i \in [0..3]$$
-
-#### Impact of the Scheme:
-Because the `nonce` increments strictly with each outgoing datagram, the session ID's masking keystream continuously changes. This breaks all packet header correlations and eliminates repeating byte patterns, rendering statistical fingerprinting futile.
+#### Impact of the Scheme
+Because the mask is keyed on both the shared secret and the packet's own ciphertext, no two packets — even consecutive ones from the same session — share a keystream, without needing an explicit counter-based scheme. This breaks all packet header correlations and eliminates repeating byte patterns, rendering statistical fingerprinting futile.
 
 ---
 
@@ -49,3 +47,14 @@ The `AdaptivePadder` calculates dynamic dummy byte quantities to append to the p
 
 - **Dynamic Distributions**: The padding algorithms emulate length profiles commonly seen in whitelisted HTTPS or real-time video streams.
 - **Encrypted Overheads**: The appended padding resides within the AEAD cipher scope. Consequently, passive observers cannot distinguish padding bytes from useful application payload, hiding the true message boundary lengths.
+
+---
+
+## Junk Packets & TCP Fragmentation
+
+OSTP does not try to impersonate a known protocol (TLS, HTTP, or otherwise) — a fingerprint-matching filter can always be updated to catch an impersonation attempt. Instead it follows a **zapret-like** approach: no recognizable header at all, plus active manipulation of packet boundaries, so there is nothing distinctive to fingerprint in the first place.
+
+- **Junk packets**: before the handshake, the client sends a configurable number (`junk_pc`) of random-size (`junk_ps`) filler datagrams. Each carries a 4-byte marker **derived from the access key** (the `junk_marker` above) rather than a fixed constant — a fixed marker would itself be a universal signature any observer could filter on across every OSTP deployment. The server derives the same per-key marker while trying candidate keys and drops matching junk silently, before it ever reaches the "unauthorized probe" logging path.
+- **TCP fragmentation** (UoT/TCP transport only): the first packet (the handshake) is split into small chunks (`frag_chunk` bytes) with short delays (`frag_sleep` ms) between writes, so DPI that inspects only the first TCP segment never sees a complete handshake to fingerprint.
+
+Both are configurable per-profile; neither is sent over plain UDP transport, where a standalone junk datagram would look exactly like a random one-off probe to the server.

docs/en/specification.md

@@ -90,11 +90,22 @@ Because the `Nonce` is unique per packet, the mask is cryptographically independ
 
 OSTP executes a Noise Protocol Framework exchange utilizing the `Noise_NNpsk0_25519_ChaChaPoly_BLAKE2s` pattern.
 
-1.  The Registration Key (`access_key`) is converted to a 32-octet strong pre-shared key (PSK) via SHA-256.
+1.  The Registration Key (`access_key`) is converted to a 32-octet strong pre-shared key (PSK) via HKDF-SHA-256.
 2.  The PSK is integrated into the state at pattern position zero, authorizing and encrypting the very first handshaking datagram.
-3.  Ephemeral Curve25519 key exchange is evaluated to synthesize autonomous symmetric keys for subsequent read/write channels.
+3.  Ephemeral Curve25519 key exchange (`ee`) is evaluated, and the two directional transport keys are taken from Noise's `Split()` over the final chaining key `ck`.
 
-The initial handshake payload includes a Unix timestamp to mitigate replay attacks. The server enforces a strict ±30-second synchronization window.
+> **Forward secrecy.** The transport keys are derived from the chaining key
+> `ck`, which absorbs the ephemeral `ee` Diffie-Hellman result. They are **not**
+> derived from the Noise handshake hash `h` — `h` only ever absorbs public
+> transcript data (ephemeral public keys and on-wire ciphertexts) and never the
+> DH secret, so keys derived from it would give an access-key holder the ability
+> to decrypt any recorded session. Deriving from `ck` binds each session to its
+> ephemeral private keys, which are discarded after the handshake: an adversary
+> who later compromises the PSK still cannot decrypt past traffic. This is a
+> wire-breaking property gated by the internal protocol version (currently 5);
+> peers on an older version derive different keys and cannot interoperate.
+
+The initial handshake payload includes a Unix timestamp to mitigate replay attacks. The server enforces a ±300-second (5-minute) synchronization window and additionally records accepted handshakes in an anti-replay set for that window.
 
 ---
 
@@ -102,16 +113,23 @@ The initial handshake payload includes a Unix timestamp to mitigate replay attac
 
 ### 7.1 Selective-Repeat ARQ
 OSTP provides reliability over UDP using a **Selective-Repeat ARQ** mechanism:
-*   The receiver maintains a reorder buffer (default: 8192 packets).
-*   Unacknowledged packets are retransmitted after an adaptive Retransmission Time Out (RTO).
-*   Acknowledgments (ACKs) are piggybacked onto outbound data frames to minimize overhead.
-*   Backpressure is applied dynamically based on the number of in-flight unacknowledged frames.
+*   The receiver maintains a reorder buffer (default: 32768 packets) for out-of-order packet reassembly.
+*   Acknowledgments use a **Cumulative + SACK** scheme: the ACK payload contains a cumulative range `(0, expected_recv_nonce - 1)` confirming all contiguous packets received, plus up to 7 additional Selective ACK ranges for non-contiguous blocks in the reorder buffer.
+*   **Rate-limited NACK:** When a gap is detected, the receiver emits a NACK for the lowest missing nonce, but no more than once per 30ms. This prevents retransmission storms under normal UDP jitter.
+*   **Retransmission:** Unacknowledged data frames are retransmitted after an adaptive Retransmission Timeout (RTO, default: 100ms) with exponential backoff (up to 64× base RTO).
+*   **Zombie Frame Eviction:** Frames exceeding `max_retries + 4` attempts are automatically dropped from the send history, preventing unbounded memory consumption and stale retransmissions.
+*   **In-flight Counting:** Backpressure is based only on retransmittable (data) frames; control frames (ACK/NACK) are excluded from the in-flight count to prevent false backpressure under high load.
+*   **Graceful Close:** The `Closing` state processes all remaining in-flight packets before transitioning to `Closed`, preventing data loss during session teardown.
 
 ### 7.2 Adaptive Padding
 To resist traffic analysis via Packet Length Analysis (PLA), OSTP pads plaintext payloads before AEAD encryption. Padding bytes are drawn from a cryptographically secure random source. The protocol supports dynamic padding boundaries up to the maximum MTU (e.g., 1400 bytes), smoothing out recognizable application traffic bursts into constant-bitrate-like streams.
 
 ### 7.3 IP Roaming
-The server supports seamless network handoffs (e.g., transitioning from Wi-Fi to cellular networks). If a packet successfully passes AEAD authentication, the server automatically binds the Session ID to the new source IP address without requiring a session restart.
+The server supports seamless network handoffs (e.g., transitioning from Wi-Fi to cellular networks). If a packet successfully passes AEAD authentication, the server automatically binds the Session ID to the new source IP address without requiring a session restart. The server maintains a rate-limited roaming scanner (50 tokens/sec) to prevent CPU exhaustion from probing attacks.
+
+### 7.4 Session Keepalive
+*   **Client-side:** Ping/Pong frames with RTT measurement are sent every 5 seconds. If no valid UDP packet is received for 60 seconds, the client initiates reconnection.
+*   **Server-side:** Sessions with no activity for 300 seconds are automatically evicted.
 
 ---
 
@@ -119,4 +137,5 @@ The server supports seamless network handoffs (e.g., transitioning from Wi-Fi to
 
 *   **Nonce Exhaustion:** The Nonce field is 64 bits. Implementations MUST terminate and re-key a session before the Nonce overflows to prevent AEAD keystream reuse.
 *   **Session Exhaustion (DoS):** Servers MUST enforce a strict cap on concurrent sessions (e.g., 1024) and silently drop handshake attempts exceeding this limit to prevent memory exhaustion attacks.
+*   **Handshake-trial CPU DoS:** Because there is no cleartext key identifier on the wire (a deliberate stealth property), a datagram from an unknown source must be trial-decrypted against every registered key. Servers MUST bound this work: OSTP caches each key's derived secrets and time-windowed junk markers (so a trial is a cheap comparison plus one AEAD attempt per key, not a fresh HKDF/HMAC), and gates the trial path behind a global token bucket (default 100/s) so a spoofed-source flood cannot force unbounded per-packet crypto. The established-session fast path and IP-roaming path are not subject to this bucket.
 *   **Header Authentication:** The header obfuscation mechanism provides privacy, not integrity. Header integrity is mathematically guaranteed by the Poly1305 Authentication Tag, which covers the entire 12-byte header as Additional Authenticated Data (AAD).

docs/relay-config-example.json

@@ -0,0 +1,41 @@
+{
+  // OSTP Relay Node Configuration
+  // Этот узел принимает соединения от клиентов, проверяет их аутентификацию
+  // и пробрасывает трафик к целевому серверу.
+  //
+  // Архитектура цепочки:
+  //   Клиент -> [Этот Relay] -> [Relay 2] -> ... -> [Target Server]
+  //
+  // Ключи синхронизируются напрямую с API Target Server каждые N секунд.
+  // Relay не знает содержимого трафика — только проверяет HMAC-подпись.
+
+  "mode": "relay",
+
+  // Адрес, на котором relay слушает входящие соединения от клиентов
+  "listen": "0.0.0.0:50000",
+
+  // Адрес следующего узла в цепочке (другой relay или конечный сервер) — TCP (UoT)
+  "upstream_tcp": "TARGET_SERVER_IP:50000",
+
+  // Адрес следующего узла в цепочке — UDP
+  "upstream_udp": "TARGET_SERVER_IP:50000",
+
+  // URL API конечного (целевого) сервера для синхронизации access_keys.
+  // Должен быть доступен с этого relay-сервера (можно через SSH-туннель).
+  //
+  // ВАЖНО: URL обязан включать секретный путь панели (api.webpath целевого
+  // сервера). Management API смонтирован ВНУТРИ этого пути — именно он скрывает
+  // панель от сканеров, — поэтому голый host:port попадает в несуществующий
+  // маршрут, и синхронизация падает с 404 ещё до проверки токена.
+  // Это тот же адрес, по которому вы открываете веб-панель.
+  "upstream_api_url": "http://TARGET_SERVER_IP:9090/TARGET_SERVER_WEBPATH",
+
+  // Bearer-токен для доступа к API целевого сервера
+  // Должен совпадать с api.token в конфиге target-сервера
+  "upstream_api_token": "YOUR_API_TOKEN_HERE",
+
+  // Интервал синхронизации ключей в секундах (по умолчанию: 30)
+  "sync_interval_secs": 30,
+
+  "debug": false
+}

docs/ru/obfuscation.md

@@ -1,51 +1,60 @@
-# Маскирование энтропии сигналов OSTP
+# Обфускация трафика OSTP
 
-## Философия структуры канала
+## Философия
 
-Традиционные сетевые протоколы промышленного сбора данных могут обладать фиксированными заголовками, что при анализе статистического распределения байт ведет к предвзятости выборок и искажению телеметрического профиля. Задача механизмов энтропийного маскирования OSTP — достижение **равномерного вероятностного распределения значений байт**, начиная с самого первого пакета. Это делает сигналы шины данных абсолютно однородными и устойчивыми к корреляционному анализу и структурному мониторингу сетевых контроллеров.
+Классические туннельные протоколы (TLS, OpenVPN, WireGuard) имеют узнаваемые сигнатуры в хэндшейке или статичные заголовки пакетов. Механизм обфускации OSTP спроектирован так, чтобы **начиная с первого байта** трафик был максимально похож на случайный шум — и для DPI-систем был неотличим от него.
 
 ---
 
-## Производная сигнатурная матрица (Keystream Initialization Vector)
+## Деривация секретов
 
-Для стабилизации битового распределения используется 8-байтовый вектор, вычисляемый на базе глобального идентификатора регистрации узла (`access_key`):
+Все секреты протокола — ключ обфускации, PSK Noise-хэндшейка, диапазон паддинга хэндшейка и маркер junk-пакетов (см. ниже) — выводятся из общего `access_key` одним проходом HKDF-SHA256, с разделением по доменам через последний байт `info`:
 
-$$\text{Key} = \text{SHA-256}(\text{access\_key})[0..8]$$
+```
+PRK              = HKDF-Extract(salt = SHA-256(access_key)[0..16], IKM = access_key || PROTOCOL_VERSION)
+obfuscation_key  = HKDF-Expand(PRK, info = SHA-256(access_key)[16..] || 0x01, 8 байт)
+psk              = HKDF-Expand(PRK, info = SHA-256(access_key)[16..] || 0x02, 32 байта)
+handshake_pad    = HKDF-Expand(PRK, info = SHA-256(access_key)[16..] || 0x03, 2 байта)
+junk_marker      = HKDF-Expand(PRK, info = SHA-256(access_key)[16..] || 0x04, 4 байта)
+```
 
-Данная последовательность фиксируется на передающем и принимающем узлах и не передается через внешние сетевые шлюзы.
+Версия протокола подмешивается в IKM, а не передаётся открытым байтом на проводе: пиры с разной версией протокола выведут разный `obfuscation_key` и просто не смогут деобфусцировать пакеты друг друга — жёсткий version gate без единого узнаваемого маркера на проводе. Ни один секрет никогда не передаётся — обе стороны независимо выводят одинаковые значения из общего access_key.
 
 ---
 
-## Алгоритм динамического маскирования пакетов (In-place Masking)
+## Алгоритм динамического маскирования
 
-Пакетные структуры OSTP проходят низкоуровневую предобработку непосредственно перед выдачей в канальный уровень (Layer 3) и при получении. В зависимости от фазы жизненного цикла сессии связи выделяют две модели:
+Датаграммы OSTP маскируются "на месте" прямо перед отправкой и сразу после получения. Сама маска **выводится из шифротекста самого пакета**, а не из статичного потока ключа или счётчика — поэтому она меняется от пакета к пакету автоматически:
 
-### 1. Этап начального согласования среды (`is_handshake = true`)
-В период инициализации канала передачи пакет структурирован как 4-байтовое поле логического адреса порта `session_id` и криптографический блок согласования среды. Для подавления статических компонент ID порта применяется процедура обратимого битового сложения:
+```
+mask = HMAC-SHA256(key = obfuscation_key, message = ciphertext[0..min(32, len)])
+```
 
-*   **Обработка**: Первые 4 байта вектора пакета проходят побитовую операцию XOR с первыми 4 байтами сигнатурной матрицы:
-    $$\text{raw}[i] = \text{raw}[i] \oplus \text{Key}[i \pmod 8], \quad i \in [0..3]$$
-*   **Восстановление**: Обратное наложение сигнатурной матрицы возвращает корректное значение логического идентификатора.
+### 1. Фаза хэндшейка (`is_handshake = true`)
+Пакет на проводе — `[4 байта session_id][2 байта noise_len][Noise-полезная нагрузка]`. Маска считается по Noise-полезной нагрузке (`raw[6..]`), и её первые 6 байт накладываются XOR'ом на `session_id || noise_len`.
 
-### 2. Этап высокоскоростного переноса данных (`is_handshake = false`)
-После перевода сессии в состояние активности кадр передачи принимает следующий вид:
-`[4 байта session_id]` + `[8 байт nonce]` + `[Полезная нагрузка блока]`
+### 2. Фаза передачи данных (`is_handshake = false`)
+Пакет на проводе — `[4 байта session_id][8 байт nonce][AEAD-шифротекст]`. Маска считается по шифротексту, и её первые 12 байт накладываются XOR'ом на `session_id || nonce`.
 
-Для максимизации дифференциальной энтропии применяется двухступенчатое динамическое взвешивание:
-
-1.  **Коррекция счетчика цикла (Nonce Correction)**: 8-байтовое значение инкрементного счетчика пакета подвергается побитовому сложению с вектором матрицы:
-    $$\text{nonce\_bytes}[i] = \text{nonce\_bytes}[i] \oplus \text{Key}[i], \quad i \in [0..7]$$
-2.  **Маскирование ID сессии**: 4-байтовое поле логического адреса маскируется с помощью переменной высокочастотной энтропии — младших 32 бит **исходного** показателя системного счетчика пакетов:
-    $$\text{session\_id\_bytes}[i] = \text{session\_id\_bytes}[i] \oplus \text{real\_nonce\_low32\_bytes}[i], \quad i \in [0..3]$$
-
-#### Статистическая устойчивость:
-Благодаря инкрементации счетчика на каждом цикле отправки, маскирующий поток (keystream) для поля `session_id` постоянно видоизменяется. Это полностью нивелирует фиксированные битовые паттерны во всем спектре UDP-датаграмм и исключает появление повторяющихся префиксов.
+#### Эффект схемы
+Поскольку маска зависит одновременно от общего секрета и от содержимого шифротекста конкретного пакета, никакие два пакета — даже два подряд идущих в одной сессии — не используют одинаковый ключевой поток, и для этого не нужна явная схема на основе счётчика. Это полностью убирает корреляции между заголовками пакетов и повторяющиеся байтовые паттерны, делая статистический фингерпринтинг бесполезным.
 
 ---
 
-## Выравнивание блоков по границам регистров (Adaptive Alignment)
+## Статистический паддинг
 
-Дополнительно к маскировке заголовков, протокол OSTP исключает возможность анализа поведения системы на основе длин пакетов данных. Модуль адаптивного заполнения (`AdaptivePadder`) рассчитывает оптимальный размер буфера выравнивания (`padding`), интегрируемый в структуру пакета до момента активации шифрующего каскада:
+Помимо маскирования заголовков, OSTP защищается от анализа длин пакетов (Traffic Length Analysis). `AdaptivePadder` вычисляет случайный размер мусорных байт, добавляемых к полезной нагрузке ещё до шифрования:
 
-- **Стратегия заполнения буферов**: Механизм анализирует текущую длину выборки телеметрии и производит масштабирование до типичных кратных длин промышленных сетей передачи данных и буферов потоковых агрегаторов.
-- **Изоляция выравнивания**: Данные заполнения помещаются внутрь защищенной области кадра. Внешние анализаторы топологии сети не способны определить внутренние границы между телеметрической нагрузкой и служебными полями выравнивания, видя только монолитный блок данных.
+- **Динамическое распределение**: длины паддинга подобраны так, чтобы напоминать профили длин обычного HTTPS-трафика или видеопотоков.
+- **Внутри шифротекста**: добавленный паддинг находится внутри области AEAD-шифрования — пассивный наблюдатель не может отличить паддинг от полезной нагрузки и не видит настоящую границу сообщения.
+
+---
+
+## Junk-пакеты и TCP-фрагментация
+
+OSTP не пытается притворяться известным протоколом (TLS, HTTP и т.п.) — фильтр по сигнатуре всегда можно обновить под конкретную имитацию. Вместо этого используется подход **в духе zapret**: никакого узнаваемого заголовка вообще, плюс активная манипуляция границами пакетов — фингерпринтить попросту нечего.
+
+- **Junk-пакеты**: перед хэндшейком клиент отправляет настраиваемое количество (`junk_pc`) мусорных датаграмм случайного размера (`junk_ps`). Каждая несёт 4-байтовый маркер, **выведенный из access_key** (тот самый `junk_marker` выше), а не фиксированную константу — константный маркер сам по себе стал бы универсальной сигнатурой для любого наблюдателя сразу по всем серверам OSTP. Сервер, перебирая кандидатов-ключей, выводит тот же маркер и тихо отбрасывает junk, не доходя до логирования «unauthorized probe».
+- **TCP-фрагментация** (только для транспорта UoT/TCP): первый пакет (хэндшейк) режется на мелкие куски (`frag_chunk` байт) с небольшими задержками (`frag_sleep` мс) между записями — DPI, анализирующий только первый TCP-сегмент, никогда не видит цельный хэндшейк для фингерпринтинга.
+
+Обе фичи настраиваются per-профиль; ни одна не применяется поверх обычного UDP-транспорта, где отдельная junk-датаграмма выглядела бы для сервера точь-в-точь как случайный одиночный проб.

docs/ru/specification.md

@@ -90,11 +90,23 @@ OSTP поддерживает **внутреннее криптографиче
 
 OSTP использует Noise Protocol Framework с паттерном `Noise_NNpsk0_25519_ChaChaPoly_BLAKE2s`.
 
-1.  Регистрационный ключ доступа (`access_key`) преобразуется в 32-байтный строгий предварительно распределенный ключ (PSK) через SHA-256.
-2.  PSK применяется на нулевой позиции паттерна, обеспечивая авторизацию и шифрование самой первой датаграммы рукопожатия (Zero-RTT авторизация).
-3.  Выполняется эфемерный обмен ключами Curve25519 для создания симметричных ключей передачи данных.
+1.  Регистрационный ключ доступа (`access_key`) преобразуется в 32-байтный строгий предварительно распределенный ключ (PSK) через HKDF-SHA-256.
+2.  PSK применяется на нулевой позиции паттерна, обеспечивая авторизацию и шифрование самой первой датаграммы рукопожатия.
+3.  Выполняется эфемерный обмен ключами Curve25519 (`ee`), и два однонаправленных транспортных ключа берутся из `Split()` протокола Noise над финальным chaining key `ck`.
 
-Первичная полезная нагрузка рукопожатия содержит Unix-отметку времени для защиты от атак повторного воспроизведения (Replay Attacks). Сервер строго контролирует окно синхронизации (±30 секунд).
+> **Прямая секретность (Forward Secrecy).** Транспортные ключи выводятся из
+> chaining key `ck`, который вбирает результат эфемерного обмена Диффи-Хеллмана
+> `ee`. Они **не** выводятся из handshake hash `h` протокола Noise: `h` вбирает
+> только публичные данные транскрипта (эфемерные публичные ключи и шифртексты с
+> провода) и никогда — сам DH-секрет, поэтому ключи, выведенные из `h`, дали бы
+> держателю PSK возможность расшифровать любую записанную сессию. Вывод из `ck`
+> привязывает каждую сессию к её эфемерным приватным ключам, которые
+> уничтожаются после рукопожатия: злоумышленник, скомпрометировавший PSK позже,
+> всё равно не сможет расшифровать прошлый трафик. Это свойство ломает
+> совместимость и защищено внутренней версией протокола (сейчас 5): узлы более
+> старой версии выводят другие ключи и не могут взаимодействовать.
+
+Первичная полезная нагрузка рукопожатия содержит Unix-отметку времени для защиты от атак повторного воспроизведения (Replay Attacks). Сервер контролирует окно синхронизации (±300 секунд, 5 минут) и дополнительно фиксирует принятые рукопожатия в множестве защиты от повтора на время этого окна.
 
 ---
 
@@ -119,4 +131,5 @@ OSTP обеспечивает надежную доставку поверх UDP
 
 *   **Исчерпание Nonce:** Поле Nonce имеет размер 64 бита. Реализации ОБЯЗАНЫ разрывать сессию до переполнения Nonce, чтобы предотвратить катастрофическое повторное использование гаммы AEAD-шифра.
 *   **DDoS и исчерпание ресурсов:** Серверы ДОЛЖНЫ применять жесткий лимит на количество одновременных сессий (например, 1024) и молча отбрасывать запросы на рукопожатие при превышении лимита, предотвращая атаки на исчерпание памяти.
+*   **CPU-DoS на пути перебора рукопожатия:** Поскольку на проводе нет открытого идентификатора ключа (намеренное свойство скрытности), датаграмму от неизвестного источника приходится пробно расшифровывать каждым зарегистрированным ключом. Серверы ОБЯЗАНЫ ограничивать эту работу: OSTP кэширует производные секреты каждого ключа и его junk-маркеры для текущего временно́го окна (поэтому одна попытка — это дешёвое сравнение плюс одна попытка AEAD на ключ, а не новые HKDF/HMAC), и ограничивает путь перебора глобальным token bucket (по умолчанию 100/с), так что флуд с подменённых адресов не может навязать неограниченную криптографию на пакет. Быстрый путь установленных сессий и путь IP-роуминга под этот лимит не попадают.
 *   **Целостность заголовка:** Механизм маскирования обеспечивает только скрытность, а не целостность. Целостность заголовков математически гарантируется 16-байтным тегом аутентификации Poly1305, который покрывает 12-байтный заголовок как присоединенные данные (AAD).

icons/sqare.svg

@@ -0,0 +1,15 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 512 512">
+  <defs>
+    <linearGradient id="g2" x1="0%" y1="0%" x2="100%" y2="100%">
+      <stop offset="0%" stop-color="#111827" />
+      <stop offset="100%" stop-color="#374151" />
+    </linearGradient>
+    <linearGradient id="g2_path" x1="0%" y1="0%" x2="100%" y2="100%">
+      <stop offset="0%" stop-color="#3B82F6" />
+      <stop offset="100%" stop-color="#14B8A6" />
+    </linearGradient>
+  </defs>
+  <rect width="512" height="512" rx="120" fill="url(#g2)" />
+  <path d="M144 256c0-61.9 50.1-112 112-112s112 50.1 112 112-50.1 112-112 112S144 317.9 144 256zm-48 0c0 88.4 71.6 160 160 160s160-71.6 160-160S344.4 96 256 96 96 167.6 96 256z" fill="url(#g2_path)"/>
+  <circle cx="256" cy="256" r="40" fill="#F59E0B" />
+</svg>

ostp-client/Cargo.toml

@@ -9,8 +9,24 @@ anyhow.workspace = true
 bytes.workspace = true
 tokio.workspace = true
 tracing.workspace = true
+tracing-subscriber = { version = "0.3", features = ["env-filter"] }
+tracing-appender = "0.2"
 ostp-core = { path = "../ostp-core" }
+ostp-tun = { path = "../ostp-tun" }
 rand.workspace = true
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0"
+json_comments = "0.2"
 portable-atomic.workspace = true
+chrono = "0.4"
+socket2 = "0.6.3"
+futures-util = "0.3.32"
+hmac = "0.12.1"
+sha2 = "0.10.8"
+base64 = "0.22.1"
+webpki-roots = "0.26"
+tun = { version = "0.8.9", features = ["async"] }
+netstack-smoltcp = "0.2.2"
+futures = "0.3.32"
+libc = "0.2.186"
+winapi = { version = "0.3.9", features = ["iphlpapi", "tcpmib", "processthreadsapi", "psapi", "handleapi", "winerror", "minwindef", "winnt", "iptypes", "ws2def"] }

ostp-client/src/app.rs

@@ -40,6 +40,9 @@ pub enum BridgeCommand {
     ToggleTunnel,
     NextProfile,
     ReloadConfig,
+    /// Triggered by Android NetworkCallback when the active network changes (WiFi→LTE, etc.).
+    /// Causes an immediate background reconnect without waiting for stall detection.
+    NetworkChanged,
     Shutdown,
 }
 
@@ -54,6 +57,12 @@ pub struct AppState {
     pub log_scroll: u16,
 }
 
+impl Default for AppState {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
 impl AppState {
     pub fn new() -> Self {
         Self {

ostp-client/src/bin/test_route.rs

@@ -0,0 +1,21 @@
+fn main() {
+    let socket = std::net::UdpSocket::bind("0.0.0.0:0").unwrap();
+    let port = socket.local_addr().unwrap().port();
+    println!("Bound UDP to port {}", port);
+    
+    if let Some(name) = ostp_client::tunnel::process_lookup::get_process_name_from_port_udp(port) {
+        println!("Found process for UDP port {}: {}", port, name);
+    } else {
+        println!("Process not found for UDP port {}", port);
+    }
+
+    let tcp_socket = std::net::TcpListener::bind("0.0.0.0:0").unwrap();
+    let tcp_port = tcp_socket.local_addr().unwrap().port();
+    println!("Bound TCP to port {}", tcp_port);
+
+    if let Some(name) = ostp_client::tunnel::process_lookup::get_process_name_from_port(tcp_port) {
+        println!("Found process for TCP port {}: {}", tcp_port, name);
+    } else {
+        println!("Process not found for TCP port {}", tcp_port);
+    }
+}

ostp-client/src/config.rs

@@ -12,13 +12,23 @@ pub struct ClientConfig {
     pub debug: bool,
     pub ostp: OstpConfig,
     pub local_proxy: LocalProxyConfig,
-    pub turn: TurnConfig,
+    #[serde(default)]
+    pub transport: TransportConfig,
     #[serde(default)]
     pub exclusions: ExclusionConfig,
     #[serde(default)]
     pub multiplex: MultiplexConfig,
+    pub dns_server: Option<String>,
+    #[serde(default = "default_tun_stack")]
+    pub tun_stack: String,
+    #[serde(default)]
+    pub kill_switch: bool,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub gui: Option<serde_json::Value>,
 }
 
+fn default_tun_stack() -> String { "system".to_string() }
+
 #[derive(Debug, Clone, Serialize, Deserialize, Default)]
 pub struct ExclusionConfig {
     #[serde(default)]
@@ -43,30 +53,80 @@ pub struct OstpConfig {
     pub access_key: String,
     pub handshake_timeout_ms: u64,
     pub io_timeout_ms: u64,
+    #[serde(default = "default_mtu")]
+    pub mtu: usize,
+    #[serde(default = "default_keepalive")]
+    pub keepalive_interval_sec: u64,
 }
 
+fn default_keepalive() -> u64 { 5 }
+
+fn default_mtu() -> usize { 1140 }
+
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct LocalProxyConfig {
     pub bind_addr: String,
     pub connect_timeout_ms: u64,
 }
 
+/// Transport layer configuration.
+/// `mode` = "udp" (default) or "uot" (UDP over TCP, no protocol mimicry —
+/// zapret-like: no recognizable header at all, not a fake TLS/HTTP shell).
 #[derive(Debug, Clone, Serialize, Deserialize)]
-pub struct TurnConfig {
-    pub enabled: bool,
-    pub server_addr: String,
-    pub username: String,
-    pub access_key: String,
+pub struct TransportConfig {
+    /// "udp" or "uot"
+    #[serde(default = "default_transport_mode")]
+    pub mode: String,
+    /// Split the first UoT/TCP packet (handshake) into tiny TCP segments to
+    /// break DPI that inspects the first packet. UoT/TCP only; ignored for UDP.
+    pub tcp_fragmentation: bool,
+    /// TCP chunk size (bytes)
+    #[serde(default = "default_frag_chunk")]
+    pub frag_chunk: usize,
+    /// TCP sleep duration between chunks (ms)
+    #[serde(default = "default_frag_sleep")]
+    pub frag_sleep: u64,
+    /// [min, max] junk packet count
+    #[serde(default = "default_junk_count")]
+    pub junk_pc: [usize; 2],
+    /// [min, max] junk packet size in bytes
+    #[serde(default = "default_junk_size")]
+    pub junk_ps: [usize; 2],
 }
 
+fn default_transport_mode() -> String { "udp".to_string() }
+fn default_frag_chunk() -> usize { 2 }
+fn default_frag_sleep() -> u64 { 2 }
+fn default_junk_count() -> [usize; 2] { [2, 5] }
+fn default_junk_size() -> [usize; 2] { [100, 1000] }
+
+impl Default for TransportConfig {
+    fn default() -> Self {
+        Self {
+            mode: default_transport_mode(),
+            tcp_fragmentation: false,
+            frag_chunk: default_frag_chunk(),
+            frag_sleep: default_frag_sleep(),
+            junk_pc: default_junk_count(),
+            junk_ps: default_junk_size(),
+        }
+    }
+}
+
+
+
+
+
 impl Default for OstpConfig {
     fn default() -> Self {
         Self {
             server_addr: "127.0.0.1:50000".to_string(),
             local_bind_addr: "0.0.0.0:0".to_string(),
             access_key: String::new(),
-            handshake_timeout_ms: 10000,
+            handshake_timeout_ms: 5000,
             io_timeout_ms: 2500,
+            mtu: default_mtu(),
+            keepalive_interval_sec: default_keepalive(),
         }
     }
 }
@@ -80,16 +140,6 @@ impl Default for LocalProxyConfig {
     }
 }
 
-impl Default for TurnConfig {
-    fn default() -> Self {
-        Self {
-            enabled: false,
-            server_addr: String::new(),
-            username: String::new(),
-            access_key: String::new(),
-        }
-    }
-}
 
 impl Default for ClientConfig {
     fn default() -> Self {
@@ -98,9 +148,13 @@ impl Default for ClientConfig {
             debug: false,
             ostp: OstpConfig::default(),
             local_proxy: LocalProxyConfig::default(),
-            turn: TurnConfig::default(),
+            transport: TransportConfig::default(),
             exclusions: ExclusionConfig::default(),
             multiplex: MultiplexConfig::default(),
+            dns_server: None,
+            tun_stack: "system".to_string(),
+            kill_switch: false,
+            gui: None,
         }
     }
 }
@@ -123,15 +177,31 @@ struct RawUnifiedConfig {
     debug: Option<bool>,
     server: Option<String>,
     access_key: Option<String>,
+    mtu: Option<usize>,
     socks5_bind: Option<String>,
     tun: Option<RawTunSection>,
     exclude: Option<RawExcludeSection>,
     mux: Option<RawMuxSection>,
+    transport: Option<RawTransportSection>,
+    gui: Option<serde_json::Value>,
+}
+
+#[derive(Debug, Deserialize)]
+struct RawTransportSection {
+    mode: Option<String>,
+    tcp_fragmentation: Option<bool>,
+    frag_chunk: Option<usize>,
+    frag_sleep: Option<u64>,
+    junk_pc: Option<[usize; 2]>,
+    junk_ps: Option<[usize; 2]>,
 }
 
 #[derive(Debug, Deserialize)]
 struct RawTunSection {
     enable: Option<bool>,
+    dns: Option<String>,
+    stack: Option<String>,
+    kill_switch: Option<bool>,
 }
 
 #[derive(Debug, Deserialize)]
@@ -147,6 +217,8 @@ struct RawMuxSection {
     sessions: Option<usize>,
 }
 
+
+
 impl ClientConfig {
     /// Hot-reload from `config.json` placed next to the running binary.
     /// Returns a new `ClientConfig` built from the unified JSON format.
@@ -157,12 +229,14 @@ impl ClientConfig {
 
         let raw = std::fs::read_to_string(&path)
             .with_context(|| format!("failed to read {}", path.display()))?;
-        let raw: RawUnifiedConfig = serde_json::from_str(&raw)
+        let mut stripped = json_comments::StripComments::new(raw.as_bytes());
+        let raw: RawUnifiedConfig = serde_json::from_reader(&mut stripped)
             .with_context(|| format!("failed to parse {}", path.display()))?;
 
         let is_tun = raw.tun.as_ref().and_then(|t| t.enable).unwrap_or(false);
         let server = raw.server.unwrap_or_else(|| "127.0.0.1:50000".to_string());
         let key = raw.access_key.unwrap_or_default();
+        let mtu = raw.mtu.unwrap_or(default_mtu());
         let socks5 = raw.socks5_bind.unwrap_or_else(|| "127.0.0.1:1088".to_string());
         let exclusions = raw.exclude.unwrap_or(RawExcludeSection {
             domains: None,
@@ -181,14 +255,23 @@ impl ClientConfig {
                 server_addr: server,
                 local_bind_addr: "0.0.0.0:0".to_string(),
                 access_key: key,
-                handshake_timeout_ms: 10000,
+                handshake_timeout_ms: 5000,
                 io_timeout_ms: 2500,
+                mtu,
+                keepalive_interval_sec: default_keepalive(),
             },
             local_proxy: LocalProxyConfig {
                 bind_addr: socks5,
                 connect_timeout_ms: 15000,
             },
-            turn: TurnConfig::default(),
+            transport: TransportConfig {
+                mode: raw.transport.as_ref().and_then(|t| t.mode.clone()).unwrap_or_else(default_transport_mode),
+                tcp_fragmentation: raw.transport.as_ref().and_then(|t| t.tcp_fragmentation).unwrap_or(false),
+                frag_chunk: raw.transport.as_ref().and_then(|t| t.frag_chunk).unwrap_or_else(default_frag_chunk),
+                frag_sleep: raw.transport.as_ref().and_then(|t| t.frag_sleep).unwrap_or_else(default_frag_sleep),
+                junk_pc: raw.transport.as_ref().and_then(|t| t.junk_pc).unwrap_or_else(default_junk_count),
+                junk_ps: raw.transport.as_ref().and_then(|t| t.junk_ps).unwrap_or_else(default_junk_size),
+            },
             exclusions: ExclusionConfig {
                 domains: exclusions.domains.unwrap_or_default(),
                 ips: exclusions.ips.unwrap_or_default(),
@@ -198,6 +281,258 @@ impl ClientConfig {
                 enabled: mux.enabled.unwrap_or(false),
                 sessions: mux.sessions.unwrap_or(1),
             },
+            dns_server: raw.tun.as_ref().and_then(|t| t.dns.clone()),
+            tun_stack: raw.tun.as_ref().and_then(|t| t.stack.clone()).unwrap_or_else(|| "system".to_string()),
+            kill_switch: raw.tun.as_ref().and_then(|t| t.kill_switch).unwrap_or(false),
+            gui: raw.gui,
         })
     }
 }
+
+// ═══════════════════════════════════════════════════════════════════════
+// On-disk config.json shapes — client, server, and relay.
+//
+// This is the ONE place these are defined. They used to be declared locally
+// inside ostp/src/main.rs (the CLI binary) with no other consumer able to
+// see them, which is exactly how ostp-client::migrate ended up working
+// against loosely-typed serde_json::Value instead of a real schema, and how
+// the CLI, the migrator, and this crate's own hot-reload path could each
+// silently drift out of sync with what a config.json actually looks like.
+// main.rs now imports these instead of re-declaring them (see the `use
+// ostp_client::config::{...}` at its top).
+//
+// These are DELIBERATELY separate from ClientConfig/OstpConfig/etc. above:
+// this section is the friendly, minimal shape a user actually edits by
+// hand; the types above are what the running engine needs internally
+// (handshake/io timeouts, resolved addresses, ...) and are built FROM one
+// of these via the mapping in ostp/src/main.rs::run_client_directly. Only
+// `ClientConfig` collides by name with the runtime type above, so the
+// on-disk one is `ClientFileConfig` — everything else keeps its natural name.
+// ═══════════════════════════════════════════════════════════════════════
+
+#[derive(Debug, Deserialize, Serialize)]
+#[serde(tag = "mode", rename_all = "lowercase")]
+pub enum AppMode {
+    Server(ServerConfig),
+    Client(ClientFileConfig),
+    Relay(RelayServerConfig),
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct UnifiedConfig {
+    #[serde(flatten)]
+    pub mode: AppMode,
+    pub log_level: Option<String>,
+}
+
+impl UnifiedConfig {
+    pub fn validate(&self) -> Result<()> {
+        match &self.mode {
+            AppMode::Server(cfg) => {
+                if cfg.access_keys.is_empty() {
+                    anyhow::bail!("Server configuration must contain at least one access_key.");
+                }
+                if let Some(outbound) = &cfg.outbound {
+                    if outbound.enabled {
+                        let action = outbound.default_action.as_deref().unwrap_or("direct");
+                        if action == "direct" && outbound.rules.is_empty() {
+                            println!("\n[WARNING] Server outbound proxy is ENABLED, but default_action is 'direct' and there are no rules!");
+                            println!("          This means ALL traffic will bypass the proxy and go out directly from the server IP.");
+                            println!("          If you want all traffic to be proxied, change 'default_action' to 'proxy'.\n");
+                        }
+                    }
+                }
+            }
+            AppMode::Client(cfg) => {
+                if cfg.access_key.is_empty() {
+                    anyhow::bail!("Client configuration must contain an access_key.");
+                }
+            }
+            AppMode::Relay(cfg) => {
+                if cfg.upstream_tcp.is_empty() {
+                    anyhow::bail!("Relay configuration must specify upstream_tcp address.");
+                }
+                if cfg.upstream_api_url.is_empty() {
+                    anyhow::bail!("Relay configuration must specify upstream_api_url.");
+                }
+            }
+        }
+        Ok(())
+    }
+}
+
+#[derive(Debug, Deserialize, Serialize, Clone)]
+#[serde(untagged)]
+pub enum UserConfig {
+    Detailed {
+        access_key: String,
+        name: Option<String>,
+        limit_bytes: Option<u64>,
+    },
+    KeyOnly(String),
+}
+
+impl UserConfig {
+    pub fn key(&self) -> String {
+        match self {
+            UserConfig::KeyOnly(k) => k.clone(),
+            UserConfig::Detailed { access_key, .. } => access_key.clone(),
+        }
+    }
+    pub fn name(&self) -> Option<String> {
+        match self {
+            UserConfig::KeyOnly(_) => None,
+            UserConfig::Detailed { name, .. } => name.clone(),
+        }
+    }
+    pub fn limit(&self) -> Option<u64> {
+        match self {
+            UserConfig::KeyOnly(_) => None,
+            UserConfig::Detailed { limit_bytes, .. } => *limit_bytes,
+        }
+    }
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct ServerConfig {
+    pub listen: ListenConfig,
+    pub access_keys: Vec<UserConfig>,
+    pub debug: Option<bool>,
+    pub outbound: Option<OutboundConfig>,
+    pub api: Option<ApiConfig>,
+    pub fallback: Option<FallbackCfg>,
+    pub transport: Option<TransportConfigRaw>,
+    // Left untyped: ostp-client does not (and should not) depend on
+    // ostp-server just to name its DnsConfig type. The CLI binary — which
+    // already depends on both crates — deserializes this into
+    // ostp_server::dns::DnsConfig right before handing it to run_server().
+    pub dns: Option<serde_json::Value>,
+}
+
+/// Relay-node config.json shape.
+#[derive(Debug, Deserialize, Serialize)]
+pub struct RelayServerConfig {
+    /// Listen address(es) (UDP + TCP UoT)
+    pub listen: ListenConfig,
+    /// Upstream address for TCP (UoT) traffic
+    pub upstream_tcp: String,
+    /// Upstream address for UDP traffic
+    pub upstream_udp: String,
+    // ── Deprecated ──────────────────────────────────────────────────────────
+    // The relay used to authenticate clients itself and pulled the access-key
+    // list from the target server's management API to do it. It no longer does:
+    // sessions are authenticated end-to-end by the target server, and a relay
+    // that re-checks credentials only adds a weaker second gate plus a copy of
+    // the key list on a machine that does not need one. These are kept solely
+    // so existing relay configs still parse; they are ignored.
+    #[serde(default)]
+    pub upstream_api_url: String,
+    #[serde(default)]
+    pub upstream_api_token: String,
+    #[serde(default)]
+    pub sync_interval_secs: u64,
+    pub debug: Option<bool>,
+}
+
+/// Supports both a single string "0.0.0.0:50000" and an array
+/// ["0.0.0.0:50000", "[::]:50000"].
+#[derive(Debug, Deserialize, Serialize, Clone)]
+#[serde(untagged)]
+pub enum ListenConfig {
+    Single(String),
+    Multiple(Vec<String>),
+}
+
+impl ListenConfig {
+    pub fn addresses(&self) -> Vec<String> {
+        match self {
+            ListenConfig::Single(s) => vec![s.clone()],
+            ListenConfig::Multiple(v) => v.clone(),
+        }
+    }
+
+    pub fn primary(&self) -> String {
+        match self {
+            ListenConfig::Single(s) => s.clone(),
+            ListenConfig::Multiple(v) => v.first().cloned().unwrap_or_default(),
+        }
+    }
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct ApiConfig {
+    pub enabled: Option<bool>,
+    pub bind: Option<String>,
+    pub token: Option<String>,
+    pub webpath: Option<String>,
+    pub username: Option<String>,
+    pub password_hash: Option<String>,
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct FallbackCfg {
+    pub enabled: Option<bool>,
+    pub listen: Option<String>,
+    pub target: Option<String>,
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct ClientFileConfig {
+    pub server: String,
+    pub access_key: String,
+    pub mtu: Option<usize>,
+    pub socks5_bind: Option<String>,
+    pub tun: Option<TunConfig>,
+    pub debug: Option<bool>,
+    pub exclude: Option<ExcludeConfig>,
+    pub mux: Option<MuxConfig>,
+    pub transport: Option<TransportConfigRaw>,
+    pub gui: Option<serde_json::Value>,
+}
+
+#[derive(Debug, Deserialize, Serialize, Clone)]
+pub struct TransportConfigRaw {
+    pub mode: Option<String>,
+    pub tcp_fragmentation: Option<bool>,
+}
+
+#[derive(Debug, Deserialize, Serialize, Clone)]
+pub struct TunConfig {
+    pub enable: bool,
+    pub wintun_path: Option<String>,
+    pub ipv4_address: Option<String>,
+    pub dns: Option<String>,
+    pub kill_switch: Option<bool>,
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct OutboundConfig {
+    pub enabled: bool,
+    pub protocol: String,
+    pub address: String,
+    pub port: u16,
+    #[serde(default)]
+    pub rules: Vec<OutboundRule>,
+    pub default_action: Option<String>,
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct OutboundRule {
+    pub domain_suffix: Option<Vec<String>>,
+    pub ip_cidr: Option<Vec<String>>,
+    pub protocol: Option<String>,
+    pub action: Option<String>,
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct ExcludeConfig {
+    pub domains: Option<Vec<String>>,
+    pub ips: Option<Vec<String>>,
+    pub processes: Option<Vec<String>>,
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct MuxConfig {
+    pub enabled: Option<bool>,
+    pub sessions: Option<usize>,
+}

ostp-client/src/lib.rs

@@ -1,7 +1,12 @@
 pub mod app;
 pub mod bridge;
 pub mod config;
+pub mod migrate;
 pub mod signal;
 pub mod sysproxy;
+pub mod transport;
 pub mod tunnel;
+
+
 pub mod runner;
+pub mod logging;

ostp-client/src/logging.rs

@@ -0,0 +1,189 @@
+use std::fs::OpenOptions;
+use std::io::Write;
+use std::path::PathBuf;
+use tracing_subscriber::{layer::SubscriberExt, util::SubscriberInitExt, EnvFilter};
+
+/// The single canonical log file for the whole core. Every process (CLI daemon,
+/// GUI, TUN helper) and every subsystem (tracing, the core event logger, the
+/// helper IPC, panics) writes here — no more per-binary / per-subsystem sprawl
+/// (`ostp-cli.log` + `ostp-core.log` + `ostp-helper.log` + `ostp-crash.log`).
+pub const LOG_FILE_NAME: &str = "ostp.log";
+
+/// Absolute path to the shared log file, next to the running executable.
+pub fn log_file_path() -> PathBuf {
+    std::env::current_exe()
+        .ok()
+        .and_then(|p| p.parent().map(|d| d.join(LOG_FILE_NAME)))
+        .unwrap_or_else(|| PathBuf::from(LOG_FILE_NAME))
+}
+
+/// True if this invocation is the long-running daemon (a client/server run),
+/// as opposed to a one-shot subcommand (`gk`, `check`, `init`, `-V`, ...).
+///
+/// Used to gate log truncation: only the daemon clears the log at startup, so a
+/// one-shot command run while a daemon is live can never wipe the daemon's log.
+/// A daemon invocation is simply one that carries none of the one-shot tokens
+/// (`ostp`, `ostp run`, `ostp connect <url>` → daemon; everything else → one-shot).
+pub fn invocation_is_daemon<I: IntoIterator<Item = String>>(args: I) -> bool {
+    const ONE_SHOT: &[&str] = &[
+        "gk", "generate-key", "check", "init", "setup", "links", "import",
+        "update", "migrate", "prober", "proxy-env", "proxy-env-clear",
+        "uninstall", "-V", "--version", "-h", "--help", "help",
+    ];
+    !args
+        .into_iter()
+        .skip(1) // program name
+        .any(|a| ONE_SHOT.contains(&a.as_str()))
+}
+
+/// Append a single timestamped line to the shared log file. Used by the manual
+/// writers (core event logger, TUN helper IPC) so their output lands in the same
+/// `ostp.log` as the tracing subscriber instead of a separate file.
+pub fn append_line(msg: &str) {
+    if let Ok(mut file) = OpenOptions::new().create(true).append(true).open(log_file_path()) {
+        let _ = writeln!(
+            file,
+            "[{}] {}",
+            chrono::Local::now().format("%Y-%m-%d %H:%M:%S"),
+            msg
+        );
+    }
+}
+
+pub fn setup_panic_hook() {
+    std::panic::set_hook(Box::new(|info| {
+        let payload = info.payload();
+        let msg = if let Some(s) = payload.downcast_ref::<&str>() {
+            *s
+        } else if let Some(s) = payload.downcast_ref::<String>() {
+            s.as_str()
+        } else {
+            "Box<dyn Any>"
+        };
+
+        let location = info.location().unwrap_or_else(|| std::panic::Location::caller());
+        let backtrace = std::backtrace::Backtrace::force_capture();
+
+        let crash_msg = format!(
+            "[{}] PANIC at {}:{}\nMessage: {}\nBacktrace:\n{:?}",
+            chrono::Local::now().format("%Y-%m-%d %H:%M:%S"),
+            location.file(),
+            location.line(),
+            msg,
+            backtrace
+        );
+
+        eprintln!("{}", crash_msg);
+        tracing::error!("{}", crash_msg);
+
+        // Crashes land in the same shared log file (append — a crash must never
+        // truncate, and the tracing worker may already be dead so we write direct).
+        if let Ok(mut file) = OpenOptions::new().create(true).append(true).open(log_file_path()) {
+            let _ = file.write_all(crash_msg.as_bytes());
+            let _ = file.write_all(b"\n===================================================\n");
+        }
+    }));
+}
+
+/// Initialises tracing and writes to the shared `ostp.log` next to the executable.
+///
+/// The `level` parameter controls the minimum log level:
+/// - `"error"` — only errors
+/// - `"warn"`  — warnings and errors
+/// - `"info"`  — informational messages (default)
+/// - `"debug"` — detailed debug messages (use when `debug: true` in config)
+/// - `"trace"` — all messages including very verbose internal state
+///
+/// The environment variable `RUST_LOG` overrides this value if set.
+///
+/// `truncate`: clear the log at startup. Honoured **only on Windows** — Linux
+/// servers keep their history (OS-rotated). Pass `true` only from the daemon's
+/// own entrypoint; one-shot commands and child processes (the TUN helper) pass
+/// `false` so they append instead of wiping a running daemon's log.
+pub fn init_tracing(
+    level: &str,
+    app_name: &str,
+    version: &str,
+    truncate: bool,
+) -> Option<tracing_appender::non_blocking::WorkerGuard> {
+    // RUST_LOG overrides the config-derived level
+    let env_filter = EnvFilter::try_from_default_env()
+        .unwrap_or_else(|_| {
+            // When debug or trace is requested, enable for all ostp crates
+            if level == "debug" || level == "trace" {
+                // Enable the requested level for ostp crates, but keep noisy deps at warn
+                EnvFilter::new(format!(
+                    "warn,ostp_client={level},ostp_core={level},ostp_jni={level},ostp_gui_lib={level}"
+                ))
+            } else {
+                EnvFilter::new(level)
+            }
+        });
+
+    let path = log_file_path();
+
+    let mut open_opts = OpenOptions::new();
+    open_opts.create(true);
+    // Truncate-on-startup is Windows-only and daemon-only. Everywhere else append:
+    // Linux keeps server history, and one-shot commands / the TUN helper must not
+    // wipe a running daemon's log.
+    if truncate && cfg!(windows) {
+        open_opts.write(true).truncate(true);
+    } else {
+        open_opts.append(true);
+    }
+
+    if let Ok(mut file) = open_opts.open(&path) {
+        // Write the startup banner directly to the log file, bypassing the
+        // tracing subscriber entirely. Emitting it via tracing::info!() hits
+        // BOTH layers below (file AND stderr), so every one-shot CLI command
+        // (`ostp -V`, `ostp gk`, `ostp check`, ...) printed this banner to the
+        // terminal on every single invocation — pure noise for anything that
+        // isn't the long-running daemon. It's still genuinely useful for
+        // whoever's reading the log file later, so keep it there, just not on
+        // screen for commands that aren't the daemon.
+        let _ = writeln!(
+            file,
+            "{} v{} | OS: {} | Arch: {} | log_level: {} | log_file: {}",
+            app_name,
+            version,
+            std::env::consts::OS,
+            std::env::consts::ARCH,
+            level,
+            path.display(),
+        );
+
+        let (file_writer, guard) = tracing_appender::non_blocking(file);
+
+        let fmt_layer = tracing_subscriber::fmt::layer()
+            .with_target(true)
+            .with_line_number(true)
+            .with_thread_ids(false)
+            .with_thread_names(false)
+            .with_ansi(false)
+            .with_writer(file_writer);
+
+        let stderr_layer = tracing_subscriber::fmt::layer()
+            .with_target(true)
+            .with_writer(std::io::stderr);
+
+        let _ = tracing_subscriber::registry()
+            .with(env_filter)
+            .with(fmt_layer)
+            .with(stderr_layer)
+            .try_init();
+
+        Some(guard)
+    } else {
+        // Fallback: stderr only
+        let stderr_layer = tracing_subscriber::fmt::layer()
+            .with_target(true)
+            .with_writer(std::io::stderr);
+        let _ = tracing_subscriber::registry()
+            .with(EnvFilter::new(level))
+            .with(stderr_layer)
+            .try_init();
+        eprintln!("[WARN] Could not open log file at {}. Logging to stderr only.", path.display());
+        None
+    }
+}

ostp-client/src/migrate.rs

@@ -0,0 +1,559 @@
+//! The ONE authoritative place that upgrades an old `config.json` to the
+//! current schema. Reachable only via the explicit `ostp migrate` command —
+//! nothing else in this codebase silently rewrites a user's config on their
+//! behalf (the old 0.3.x line used to auto-migrate on every load with just a
+//! log warning; that's exactly the kind of "invisible until something looks
+//! wrong" behavior this module replaces).
+//!
+//! Every field this module cannot map forward is reported explicitly in
+//! `MigrationReport.notes`, never silently dropped without a trace.
+
+use serde_json::{json, Value};
+
+#[derive(Debug, Default)]
+pub struct MigrationReport {
+    /// Whether anything was actually different from the current schema.
+    pub changed: bool,
+    /// Human-readable line per field added, converted, or dropped.
+    pub notes: Vec<String>,
+}
+
+impl MigrationReport {
+    fn note(&mut self, msg: impl Into<String>) {
+        self.changed = true;
+        self.notes.push(msg.into());
+    }
+}
+
+/// Which config this file is (mirrors `AppMode`'s `"mode"` tag). Old configs
+/// from before that tag existed are sniffed structurally as a fallback.
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum ConfigKind {
+    Client,
+    Server,
+    Relay,
+}
+
+pub fn detect_kind(json: &Value) -> Option<ConfigKind> {
+    match json.get("mode").and_then(|v| v.as_str()) {
+        Some("client") => return Some(ConfigKind::Client),
+        Some("server") => return Some(ConfigKind::Server),
+        Some("relay") => return Some(ConfigKind::Relay),
+        _ => {}
+    }
+    // No (or unrecognized) "mode" tag — this is an older config from before
+    // it was mandatory. Sniff by the fields that have been present on each
+    // shape since the earliest surviving config format.
+    if json.get("upstream_tcp").is_some() || json.get("upstream_api_url").is_some() {
+        Some(ConfigKind::Relay)
+    } else if json.get("access_keys").is_some() || json.get("listen").is_some() {
+        Some(ConfigKind::Server)
+    } else if json.get("access_key").is_some() || json.get("server").is_some() {
+        Some(ConfigKind::Client)
+    } else {
+        None
+    }
+}
+
+/// Migrates a client config of any known past shape to the current flat
+/// schema. Returns the migrated JSON and a report of every change made.
+///
+/// Known input shapes, oldest first:
+/// - **v0.3.1–v0.3.21 "modular multi-server"**: `inbounds`/`outbounds` arrays
+///   + `routing.rules`. Only the first `ostp`-type outbound is kept (this
+///   line no longer supports multiple simultaneous servers); every other
+///   `ostp` outbound is reported by tag+address so nothing vanishes
+///   invisibly. `urltest`/`selector`/`direct`/`block` outbounds have no
+///   equivalent and are dropped (reported).
+/// - **pre-0.3.1 flat (up to v0.2.98)**: same field names as today
+///   (`server`, `access_key`, `tun`, `exclude`, `mux`, `transport`, ...)
+///   except `tun.wintun_path`/`tun.ipv4_address` (internal driver detail,
+///   never user-meaningful data) and `transport.wss` (the WSS framing
+///   feature removed entirely in the 0.4.0 rebuild) — both dropped with an
+///   explicit note; everything else maps 1:1, nothing to convert.
+/// - **configs carrying a leftover `transport.stealth_sni`**: dropped with a
+///   note, same reasoning as `wss` — it never fed into anything on the wire
+///   (no TLS/HTTP mimicry exists in this project), so there is no successor
+///   field. Not tied to a specific version: it lingered in the schema well
+///   past when the mimicry work it was meant for got removed.
+/// - **current flat schema**: no-op, `changed = false`.
+pub fn migrate_client_json(json: Value) -> (Value, MigrationReport) {
+    let mut report = MigrationReport::default();
+
+    let has_inbounds = json.get("inbounds").and_then(|v| v.as_array()).is_some();
+    let has_outbounds = json.get("outbounds").and_then(|v| v.as_array()).is_some();
+
+    if has_inbounds && has_outbounds {
+        return migrate_client_from_modular(json, report);
+    }
+
+    // Flat shape already (current or pre-0.3.1) — normalize obsolete fields
+    // in place rather than rebuilding the whole document from scratch, so
+    // any field this module doesn't know about yet still survives untouched.
+    let mut out = json;
+
+    if let Some(tun) = out.get_mut("tun").and_then(|t| t.as_object_mut()) {
+        for dead_field in ["wintun_path", "ipv4_address"] {
+            if tun.remove(dead_field).is_some() {
+                report.note(format!(
+                    "Dropped tun.{dead_field} — internal driver detail from an older WinTun \
+                     integration, not applicable to the current TUN implementation."
+                ));
+            }
+        }
+    }
+    if let Some(transport) = out.get_mut("transport").and_then(|t| t.as_object_mut()) {
+        if transport.remove("wss").is_some() {
+            report.note(
+                "Dropped transport.wss — WSS framing was removed in the 0.4.0 rebuild \
+                 (the project follows a zapret-like approach: no protocol mimicry, \
+                 just packet-level obfuscation/manipulation, so there is no successor field)."
+                    .to_string(),
+            );
+        }
+        if transport.remove("stealth_sni").is_some() {
+            report.note(
+                "Dropped transport.stealth_sni — never actually used to construct any wire \
+                 bytes (no TLS/HTTP mimicry exists in this project — same zapret-like \
+                 reasoning as transport.wss), so it was unused config plumbing with no effect."
+                    .to_string(),
+            );
+        }
+    }
+
+    (out, report)
+}
+
+fn migrate_client_from_modular(json: Value, mut report: MigrationReport) -> (Value, MigrationReport) {
+    report.changed = true; // the shape itself is being replaced regardless of field-level detail
+
+    let inbounds = json.get("inbounds").and_then(|v| v.as_array()).cloned().unwrap_or_default();
+    let outbounds = json.get("outbounds").and_then(|v| v.as_array()).cloned().unwrap_or_default();
+    let routing = json.get("routing").cloned().unwrap_or(json!({}));
+    let default_outbound = routing.get("default_outbound").and_then(|v| v.as_str()).map(String::from);
+
+    // ── Pick the primary "ostp" outbound ────────────────────────────────
+    // Prefer the one routing.default_outbound points at (directly, or via a
+    // urltest/selector group that references it); otherwise take the first
+    // ostp outbound in file order. Every other ostp outbound is reported by
+    // tag+address, not silently discarded.
+    let ostp_outbounds: Vec<&Value> = outbounds
+        .iter()
+        .filter(|o| o.get("type").and_then(|t| t.as_str()) == Some("ostp"))
+        .collect();
+
+    // default_outbound might name an ostp outbound directly, OR name a
+    // urltest/selector GROUP whose first member is the one to actually use —
+    // check both, since a plain `.or_else` here would never even attempt the
+    // group lookup while default_outbound is Some(_) (which it almost always
+    // is), silently falling through to "just take the first ostp outbound in
+    // file order" instead — exactly the kind of silent wrong answer this
+    // migrator exists to avoid.
+    let primary_tag: Option<String> = default_outbound.as_deref().and_then(|def_tag| {
+        if ostp_outbounds.iter().any(|o| o.get("tag").and_then(|t| t.as_str()) == Some(def_tag)) {
+            return Some(def_tag.to_string());
+        }
+        outbounds.iter().find_map(|o| {
+            let is_group = matches!(o.get("type").and_then(|t| t.as_str()), Some("urltest") | Some("selector"));
+            let tag_matches = o.get("tag").and_then(|t| t.as_str()) == Some(def_tag);
+            if is_group && tag_matches {
+                o.get("outbounds")
+                    .and_then(|v| v.as_array())
+                    .and_then(|arr| arr.first())
+                    .and_then(|v| v.as_str())
+                    .map(String::from)
+            } else {
+                None
+            }
+        })
+    });
+
+    let primary = primary_tag
+        .as_deref()
+        .and_then(|tag| ostp_outbounds.iter().find(|o| o.get("tag").and_then(|t| t.as_str()) == Some(tag)))
+        .copied()
+        .or_else(|| ostp_outbounds.first().copied());
+
+    let Some(primary) = primary else {
+        report.note(
+            "No 'ostp'-type outbound found in the old modular config — nothing to migrate \
+             the server connection from. Wrote a placeholder; you MUST fill in server/access_key \
+             by hand or re-import a share link."
+                .to_string(),
+        );
+        return (
+            json!({
+                "server": "127.0.0.1:50000",
+                "access_key": "",
+            }),
+            report,
+        );
+    };
+
+    for other in &ostp_outbounds {
+        if !std::ptr::eq(*other, primary) {
+            let tag = other.get("tag").and_then(|t| t.as_str()).unwrap_or("?");
+            let addr = other.get("server").and_then(|t| t.as_str()).unwrap_or("?");
+            let port = other.get("port").and_then(|t| t.as_u64()).unwrap_or(0);
+            report.note(format!(
+                "Dropped additional server '{tag}' ({addr}:{port}) — multi-server / urltest \
+                 failover is no longer supported; only one server per config now. Kept the \
+                 one from routing.default_outbound (or the first one if that wasn't set)."
+            ));
+        }
+    }
+
+    let server = primary.get("server").and_then(|v| v.as_str()).unwrap_or("127.0.0.1").to_string();
+    let port = primary.get("port").and_then(|v| v.as_u64()).unwrap_or(50000);
+    let access_key = primary.get("access_key").and_then(|v| v.as_str()).unwrap_or("").to_string();
+    let transport_type = primary
+        .get("transport")
+        .and_then(|t| t.get("type").or_else(|| t.get("mode")))
+        .and_then(|v| v.as_str())
+        .unwrap_or("udp")
+        .to_string();
+    if let Some(sni) = primary.get("transport").and_then(|t| t.get("stealth_sni")).and_then(|v| v.as_str()) {
+        if !sni.is_empty() {
+            report.note(format!(
+                "Dropped transport.stealth_sni ({sni:?}) — never actually used to construct \
+                 any wire bytes; unused config plumbing with no successor field."
+            ));
+        }
+    }
+    let tcp_fragmentation = primary
+        .get("transport")
+        .and_then(|t| t.get("tcp_fragmentation"))
+        .and_then(|v| v.as_bool())
+        .unwrap_or(false);
+    let mux_enabled = primary.get("multiplex").and_then(|m| m.get("enabled")).and_then(|v| v.as_bool()).unwrap_or(false);
+    let mux_sessions = primary.get("multiplex").and_then(|m| m.get("sessions")).and_then(|v| v.as_u64()).unwrap_or(1);
+
+    // ── TUN + local proxy inbounds ───────────────────────────────────────
+    let tun_inbound = inbounds.iter().find(|i| i.get("type").and_then(|t| t.as_str()) == Some("tun"));
+    let proxy_inbound = inbounds.iter().find(|i| i.get("type").and_then(|t| t.as_str()) == Some("local_proxy"));
+
+    let tun_enable = tun_inbound.is_some();
+    let mtu = tun_inbound.and_then(|t| t.get("mtu")).and_then(|v| v.as_u64());
+
+    let socks5_bind = proxy_inbound
+        .map(|p| {
+            let listen = p.get("listen").and_then(|v| v.as_str()).unwrap_or("127.0.0.1");
+            let port = p.get("port").and_then(|v| v.as_u64()).unwrap_or(1088);
+            format!("{listen}:{port}")
+        })
+        .unwrap_or_else(|| "127.0.0.1:1088".to_string());
+
+    // ── Exclusions from routing.rules → direct ──────────────────────────
+    let mut ex_domains: Vec<String> = Vec::new();
+    let mut ex_ips: Vec<String> = Vec::new();
+    let mut ex_processes: Vec<String> = Vec::new();
+    if let Some(rules) = routing.get("rules").and_then(|v| v.as_array()) {
+        for rule in rules {
+            if rule.get("outbound").and_then(|v| v.as_str()) != Some("direct") {
+                continue; // only "route to direct" rules were ever exclusions in the old format
+            }
+            if let Some(v) = rule.get("domain_suffix").and_then(|v| v.as_array()) {
+                ex_domains.extend(v.iter().filter_map(|s| s.as_str().map(String::from)));
+            }
+            if let Some(v) = rule.get("ip_cidr").and_then(|v| v.as_array()) {
+                ex_ips.extend(v.iter().filter_map(|s| s.as_str().map(String::from)));
+            }
+            if let Some(v) = rule.get("process_name").and_then(|v| v.as_array()) {
+                ex_processes.extend(v.iter().filter_map(|s| s.as_str().map(String::from)));
+            }
+        }
+    }
+    for other_rule_outbound in routing
+        .get("rules")
+        .and_then(|v| v.as_array())
+        .into_iter()
+        .flatten()
+        .filter_map(|r| r.get("outbound").and_then(|v| v.as_str()))
+        .filter(|o| *o != "direct")
+    {
+        report.note(format!(
+            "Dropped a routing rule targeting outbound '{other_rule_outbound}' — only \
+             \"route to direct\" rules map to today's exclusions; anything else \
+             (custom per-domain outbound selection) has no equivalent anymore."
+        ));
+    }
+
+    let debug = json.get("log").and_then(|l| l.get("level")).and_then(|v| v.as_str()) == Some("debug");
+
+    let mut client = json!({
+        "server": server,
+        "port": port,
+        "access_key": access_key,
+        "socks5_bind": socks5_bind,
+        "debug": debug,
+        "tun": {
+            "enable": tun_enable,
+            "dns": null,
+            "kill_switch": false,
+        },
+        "exclude": {
+            "domains": ex_domains,
+            "ips": ex_ips,
+            "processes": ex_processes,
+        },
+        "mux": {
+            "enabled": mux_enabled,
+            "sessions": mux_sessions,
+        },
+        "transport": {
+            "mode": transport_type,
+            "tcp_fragmentation": tcp_fragmentation,
+        },
+    });
+    if let Some(mtu) = mtu {
+        client["mtu"] = json!(mtu);
+    }
+    if let Some(gui) = json.get("gui") {
+        client["gui"] = gui.clone();
+    }
+
+    (client, report)
+}
+
+/// Migrates a server config. The server shape has stayed structurally
+/// identical since the earliest surviving version — this only backfills the
+/// `api` section (added after some configs already existed) and drops the
+/// legacy `api.token` field. Ported from the ad-hoc Python snippet that used
+/// to live in `scripts/install.sh` and only ran at install/update time.
+pub fn migrate_server_json(json: Value) -> (Value, MigrationReport) {
+    let mut report = MigrationReport::default();
+    let mut out = json;
+
+    let obj = match out.as_object_mut() {
+        Some(o) => o,
+        None => return (out, report),
+    };
+
+    let api = obj.entry("api").or_insert_with(|| json!({}));
+    if let Some(api_obj) = api.as_object_mut() {
+        let defaults: [(&str, Value); 5] = [
+            ("enabled", json!(false)),
+            ("bind", json!("0.0.0.0:9090")),
+            ("webpath", json!("")),
+            ("username", json!("")),
+            ("password_hash", json!("")),
+        ];
+        for (key, default) in defaults {
+            if !api_obj.contains_key(key) {
+                report.note(format!("Added api.{key} = {default} (missing default)"));
+                api_obj.insert(key.to_string(), default);
+            }
+        }
+        if api_obj.remove("token").is_some() {
+            report.note(
+                "Dropped legacy api.token — superseded by api.password_hash; \
+                 set a new admin password with the management API or panel."
+                    .to_string(),
+            );
+        }
+    }
+
+    (out, report)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    /// A realistic v0.3.21-shaped modular config (TUN + local_proxy inbounds,
+    /// a single ostp outbound, exclusion rules, mux) — mirrors the actual
+    /// shape from that tag, field for field.
+    #[test]
+    fn modular_single_server_preserves_every_field() {
+        let old = json!({
+            "version": "0.3.21",
+            "log": { "level": "debug" },
+            "inbounds": [
+                { "type": "tun", "tag": "tun-in", "auto_route": true, "mtu": 1350 },
+                { "type": "local_proxy", "tag": "socks-in", "protocol": "socks", "listen": "127.0.0.1", "port": 1088 }
+            ],
+            "outbounds": [
+                {
+                    "type": "ostp", "tag": "proxy",
+                    "server": "203.0.113.5", "port": 50000, "access_key": "sekrit123",
+                    "transport": { "type": "uot", "stealth_sni": "vk.com", "tcp_fragmentation": true },
+                    "multiplex": { "enabled": true, "sessions": 4 }
+                },
+                { "type": "direct", "tag": "direct" },
+                { "type": "block", "tag": "block" }
+            ],
+            "routing": {
+                "rules": [
+                    { "domain_suffix": ["local.lan", "internal.corp"], "outbound": "direct" },
+                    { "ip_cidr": ["192.168.0.0/16"], "outbound": "direct" },
+                    { "process_name": ["steam.exe"], "outbound": "direct" }
+                ],
+                "default_outbound": "proxy"
+            }
+        });
+
+        let (new, report) = migrate_client_json(old);
+        assert!(report.changed);
+        assert_eq!(new["server"], "203.0.113.5");
+        assert_eq!(new["port"], 50000);
+        assert_eq!(new["access_key"], "sekrit123");
+        assert_eq!(new["socks5_bind"], "127.0.0.1:1088");
+        assert_eq!(new["mtu"], 1350);
+        assert_eq!(new["debug"], true);
+        assert_eq!(new["tun"]["enable"], true);
+        assert_eq!(new["transport"]["mode"], "uot");
+        assert_eq!(new["transport"]["tcp_fragmentation"], true);
+        assert_eq!(new["mux"]["enabled"], true);
+        assert_eq!(new["mux"]["sessions"], 4);
+        assert_eq!(new["exclude"]["domains"], json!(["local.lan", "internal.corp"]));
+        assert_eq!(new["exclude"]["ips"], json!(["192.168.0.0/16"]));
+        assert_eq!(new["exclude"]["processes"], json!(["steam.exe"]));
+        // stealth_sni never fed into any wire bytes — dropped, not carried forward.
+        assert!(new["transport"].get("stealth_sni").is_none());
+        assert!(report.notes.iter().any(|n| n.contains("stealth_sni") && n.contains("vk.com")));
+    }
+
+    /// Old modular configs that had MULTIPLE ostp outbounds (multi-server) —
+    /// must keep the one routing.default_outbound points at and report every
+    /// other one by name/address rather than picking silently.
+    #[test]
+    fn modular_multi_server_keeps_default_and_reports_the_rest() {
+        let old = json!({
+            "inbounds": [],
+            "outbounds": [
+                { "type": "ostp", "tag": "proxy-0", "server": "1.1.1.1", "port": 50000, "access_key": "k1" },
+                { "type": "ostp", "tag": "proxy-1", "server": "2.2.2.2", "port": 50000, "access_key": "k2" },
+                {
+                    "type": "urltest", "tag": "proxy",
+                    "outbounds": ["proxy-1", "proxy-0"], "url": "http://cp.cloudflare.com"
+                }
+            ],
+            "routing": { "rules": [], "default_outbound": "proxy" }
+        });
+
+        let (new, report) = migrate_client_json(old);
+        // urltest's first member (proxy-1 / 2.2.2.2) is the one actually picked.
+        assert_eq!(new["server"], "2.2.2.2");
+        assert_eq!(new["access_key"], "k2");
+        assert!(report.notes.iter().any(|n| n.contains("proxy-0") && n.contains("1.1.1.1")));
+    }
+
+    /// Pre-0.3.1 flat config carrying fields that no longer exist
+    /// (tun.wintun_path, tun.ipv4_address, transport.wss, transport.stealth_sni)
+    /// — those get dropped with a note; every field that's still meaningful
+    /// passes through untouched, byte for byte.
+    #[test]
+    fn flat_legacy_drops_only_dead_fields() {
+        let old = json!({
+            "server": "198.51.100.9:50000",
+            "access_key": "oldkey",
+            "mtu": 1200,
+            "socks5_bind": "127.0.0.1:1090",
+            "tun": {
+                "enable": true,
+                "wintun_path": "C:\\Program Files\\wintun\\wintun.dll",
+                "ipv4_address": "10.0.0.2",
+                "dns": "1.1.1.1",
+                "kill_switch": true
+            },
+            "exclude": { "domains": ["a.com"], "ips": null, "processes": null },
+            "mux": { "enabled": false, "sessions": 1 },
+            "transport": { "mode": "udp", "stealth_sni": "bing.com", "wss": true }
+        });
+
+        let (new, report) = migrate_client_json(old);
+        assert!(report.changed);
+        // Untouched fields survive exactly as they were.
+        assert_eq!(new["server"], "198.51.100.9:50000");
+        assert_eq!(new["access_key"], "oldkey");
+        assert_eq!(new["mtu"], 1200);
+        assert_eq!(new["tun"]["enable"], true);
+        assert_eq!(new["tun"]["dns"], "1.1.1.1");
+        assert_eq!(new["tun"]["kill_switch"], true);
+        assert_eq!(new["exclude"]["domains"], json!(["a.com"]));
+        // Dead fields are gone...
+        assert!(new["tun"].get("wintun_path").is_none());
+        assert!(new["tun"].get("ipv4_address").is_none());
+        assert!(new["transport"].get("wss").is_none());
+        assert!(new["transport"].get("stealth_sni").is_none());
+        // ...and their removal was reported, not silent.
+        assert!(report.notes.iter().any(|n| n.contains("wintun_path")));
+        assert!(report.notes.iter().any(|n| n.contains("ipv4_address")));
+        assert!(report.notes.iter().any(|n| n.contains("wss")));
+        assert!(report.notes.iter().any(|n| n.contains("stealth_sni")));
+    }
+
+    /// A config already in the current shape must be a true no-op: report
+    /// says nothing changed, and every field is untouched.
+    #[test]
+    fn current_flat_config_is_a_no_op() {
+        let current = json!({
+            "server": "example.com:50000",
+            "access_key": "k",
+            "tun": { "enable": false, "dns": null, "kill_switch": false },
+            "exclude": { "domains": [], "ips": [], "processes": [] },
+            "mux": { "enabled": false, "sessions": 1 },
+            "transport": { "mode": "udp", "tcp_fragmentation": false }
+        });
+        let (new, report) = migrate_client_json(current.clone());
+        assert!(!report.changed);
+        assert_eq!(new, current);
+    }
+
+    /// Every migrated output must actually deserialize into the ONE
+    /// canonical schema (`crate::config`) — this is the same check
+    /// `cmd_migrate` runs at runtime before ever touching a user's file,
+    /// exercised here directly so a schema/migrator drift fails a fast unit
+    /// test instead of surfacing as "your migrated config won't load".
+    #[test]
+    fn every_migrated_output_matches_the_canonical_schema() {
+        let modular = json!({
+            "inbounds": [{ "type": "tun", "tag": "tun-in", "mtu": 1350 }],
+            "outbounds": [
+                { "type": "ostp", "tag": "proxy", "server": "1.2.3.4", "port": 50000, "access_key": "k" },
+                { "type": "direct", "tag": "direct" }
+            ],
+            "routing": { "rules": [], "default_outbound": "proxy" }
+        });
+        let (new, _) = migrate_client_json(modular);
+        serde_json::from_value::<crate::config::ClientFileConfig>(new)
+            .expect("modular->flat migration output must match ClientFileConfig");
+
+        let legacy_flat = json!({
+            "server": "1.2.3.4:50000",
+            "access_key": "k",
+            "tun": { "enable": true, "wintun_path": "x", "ipv4_address": "y" }
+        });
+        let (new, _) = migrate_client_json(legacy_flat);
+        serde_json::from_value::<crate::config::ClientFileConfig>(new)
+            .expect("legacy-flat migration output must match ClientFileConfig");
+
+        let server = json!({ "listen": "0.0.0.0:50000", "access_keys": ["k"] });
+        let (new, _) = migrate_server_json(server);
+        serde_json::from_value::<crate::config::ServerConfig>(new)
+            .expect("server migration output must match ServerConfig");
+    }
+
+    #[test]
+    fn server_config_backfills_api_defaults_and_drops_legacy_token() {
+        let old = json!({
+            "listen": "0.0.0.0:50000",
+            "access_keys": ["k1"],
+            "api": { "token": "old-plain-token" }
+        });
+        let (new, report) = migrate_server_json(old);
+        assert!(report.changed);
+        assert_eq!(new["api"]["enabled"], false);
+        assert_eq!(new["api"]["bind"], "0.0.0.0:9090");
+        assert!(new["api"].get("token").is_none());
+        assert!(report.notes.iter().any(|n| n.contains("api.token")));
+    }
+
+    #[test]
+    fn detect_kind_falls_back_to_structural_sniffing_without_mode_tag() {
+        assert_eq!(detect_kind(&json!({"access_key": "x", "server": "y"})), Some(ConfigKind::Client));
+        assert_eq!(detect_kind(&json!({"access_keys": ["x"], "listen": "y"})), Some(ConfigKind::Server));
+        assert_eq!(detect_kind(&json!({"upstream_tcp": "x", "upstream_api_url": "y"})), Some(ConfigKind::Relay));
+        assert_eq!(detect_kind(&json!({"mode": "client", "server": "x"})), Some(ConfigKind::Client));
+    }
+}

ostp-client/src/runner.rs

@@ -6,6 +6,17 @@ use crate::bridge::{Bridge, BridgeMetrics};
 use crate::signal::wait_for_shutdown_signal;
 use crate::tunnel;
 use std::sync::Arc;
+use std::fs::OpenOptions;
+use std::io::Write as _;
+
+fn log_to_core_file(msg: &str) {
+    // Writes into the single shared ostp.log (same file as the tracing appender),
+    // not a separate ostp-core.log — see logging::LOG_FILE_NAME.
+    let path = crate::logging::log_file_path();
+    if let Ok(mut file) = OpenOptions::new().create(true).append(true).open(path) {
+        let _ = writeln!(file, "[{}] {}", chrono::Local::now().format("%Y-%m-%d %H:%M:%S"), msg);
+    }
+}
 
 #[cfg(target_os = "windows")]
 #[link(name = "kernel32")]
@@ -32,7 +43,7 @@ fn hide_console() {
 }
 
 #[cfg(target_os = "windows")]
-fn is_admin() -> bool {
+pub fn is_admin() -> bool {
     std::process::Command::new("net")
         .arg("session")
         .stdout(std::process::Stdio::null())
@@ -99,40 +110,216 @@ fn relaunch_as_admin() -> Result<()> {
     std::process::exit(0);
 }
 
+#[cfg(target_os = "linux")]
+pub fn is_root() -> bool {
+    unsafe { libc::geteuid() == 0 }
+}
+
+#[cfg(target_os = "linux")]
+fn relaunch_as_root() -> Result<()> {
+    use std::io::IsTerminal;
+    let exe = std::env::current_exe()?;
+    let args: Vec<String> = std::env::args().skip(1).collect();
+
+    let is_gui = std::env::var("DISPLAY").is_ok() || std::env::var("WAYLAND_DISPLAY").is_ok();
+    let is_term = std::io::stdout().is_terminal();
+
+    let mut cmd = if is_gui && !is_term {
+        let mut c = std::process::Command::new("pkexec");
+        c.arg(exe);
+        c
+    } else {
+        let mut c = std::process::Command::new("sudo");
+        c.arg(exe);
+        c
+    };
+
+    cmd.args(&args);
+    
+    let status = cmd.status().map_err(|e| anyhow::anyhow!("Failed to execute privilege escalation command: {}", e))?;
+    
+    if !status.success() {
+        return Err(anyhow::anyhow!("Privilege escalation failed or was denied."));
+    }
+
+    std::process::exit(0);
+}
+
 pub async fn run_client(config: crate::config::ClientConfig) -> Result<()> {
+    #[cfg(target_os = "windows")]
+    if config.mode == "tun" && !is_admin() {
+        println!("[ostp] TUN mode requires administrator privileges. Relaunching...");
+        relaunch_as_admin()?;
+    }
+
+    #[cfg(target_os = "linux")]
+    if config.mode == "tun" && !is_root() {
+        println!("[ostp] TUN mode requires root privileges. Requesting sudo/pkexec elevation...");
+        relaunch_as_root()?;
+    }
+
     let bg = std::env::args().any(|a| a == "--bg");
 
     if bg {
         hide_console();
     }
 
-    #[cfg(target_os = "windows")]
-    if config.mode == "tun" && !is_admin() {
-        println!("[ostp-client] TUN mode requires Administrator privileges. Relaunching as Admin...");
-        relaunch_as_admin()?;
-    }
-
-    if config.mode == "tun" && !config.exclusions.processes.is_empty() {
-        println!("[ostp-client] WARNING: process exclusions are not supported in the current TUN implementation");
-    }
-
-
-    let (proxy_events_tx, proxy_events_rx) = mpsc::channel(10000);
-    let (client_msgs_tx, client_msgs_rx) = mpsc::channel(10000);
-
     let metrics = Arc::new(BridgeMetrics {
         bytes_sent: portable_atomic::AtomicU64::new(0),
         bytes_recv: portable_atomic::AtomicU64::new(0),
+        connection_state: portable_atomic::AtomicU8::new(0),
+        rtt_ms: portable_atomic::AtomicU32::new(0),
     });
 
-    let bridge = Bridge::new(&config, metrics)?;
+    let (shutdown_tx, shutdown_rx) = watch::channel(false);
+
+    tokio::spawn(async move {
+        if wait_for_shutdown_signal().await.is_ok() {
+            let _ = shutdown_tx.send(true);
+        }
+    });
+
+    run_client_core(config, metrics, shutdown_rx, None).await
+}
+
+/// Runs the client with auto-reconnect: any subsystem ending — a network
+/// change stranding the TUN adapter/UDP socket on a dead interface, the OSTP
+/// protocol connection dropping in a way the inner Bridge-level retry (see
+/// `UiEvent::TunnelStopped` below) couldn't recover from, or a proxy/TUN task
+/// crashing outright — triggers a full clean restart (fresh DNS resolution,
+/// fresh Bridge, fresh TUN/proxy) with exponential backoff, instead of the
+/// client just dying. Only an explicit shutdown request stops this loop.
+pub async fn run_client_core(
+    config: crate::config::ClientConfig,
+    metrics: Arc<BridgeMetrics>,
+    mut shutdown_rx_ext: watch::Receiver<bool>,
+    config_rx: Option<watch::Receiver<crate::config::ClientConfig>>,
+) -> Result<()> {
+    use portable_atomic::Ordering;
+
+    const BACKOFF_SCHEDULE_SECS: [u64; 6] = [1, 2, 5, 10, 20, 30];
+    // A run that stayed up at least this long counts as "was actually
+    // connected", so a later drop restarts the backoff from the top instead
+    // of inheriting a long delay from a previous flaky stretch.
+    const STABLE_UPTIME: std::time::Duration = std::time::Duration::from_secs(60);
+    let mut backoff_idx = 0usize;
+
+    loop {
+        if *shutdown_rx_ext.borrow() {
+            return Ok(());
+        }
+
+        let attempt_start = std::time::Instant::now();
+        let result = run_client_once(config.clone(), metrics.clone(), shutdown_rx_ext.clone(), config_rx.clone()).await;
+
+        if *shutdown_rx_ext.borrow() {
+            // Shutdown was requested during (or right after) this attempt — honor it, don't retry.
+            return result;
+        }
+        if let Err(ref e) = result {
+            tracing::warn!("client run ended unexpectedly, will auto-reconnect: {e}");
+        }
+
+        if attempt_start.elapsed() >= STABLE_UPTIME {
+            backoff_idx = 0;
+        }
+        let delay = BACKOFF_SCHEDULE_SECS[backoff_idx.min(BACKOFF_SCHEDULE_SECS.len() - 1)];
+        backoff_idx += 1;
+
+        // Reflect the retry wait as "connecting" rather than "disconnected".
+        metrics.connection_state.store(1, Ordering::Relaxed);
+        tokio::select! {
+            _ = tokio::time::sleep(std::time::Duration::from_secs(delay)) => {}
+            _ = shutdown_rx_ext.changed() => {
+                if *shutdown_rx_ext.borrow() {
+                    return Ok(());
+                }
+            }
+        }
+    }
+}
+
+async fn run_client_once(
+    mut config: crate::config::ClientConfig,
+    metrics: Arc<BridgeMetrics>,
+    mut shutdown_rx_ext: watch::Receiver<bool>,
+    mut config_rx: Option<watch::Receiver<crate::config::ClientConfig>>,
+) -> Result<()> {
+    #[cfg(target_os = "windows")]
+    if config.mode == "tun" && !is_admin() {
+        return Err(anyhow::anyhow!("Administrator privileges are required to initialize TUN mode. Please run the application as Administrator."));
+    }
+
+    #[cfg(target_os = "linux")]
+    if config.mode == "tun" && !is_root() {
+        return Err(anyhow::anyhow!("Root privileges are required to initialize TUN mode on Linux. Please run with sudo."));
+    }
+
+    log_to_core_file(&format!("[core] Starting run_client_core in mode: {}", config.mode));
+
+    // Resolve the server IP before we override system routing and DNS.
+    // This prevents DNS deadlock if the VPN disconnects and tries to reconnect,
+    // and also ensures we add the direct route to the exact IP the bridge connects to.
+    #[allow(unused_mut)]
+    let mut resolved_addrs: Vec<std::net::SocketAddr> = tokio::net::lookup_host(&config.ostp.server_addr)
+        .await
+        .map_err(|e| anyhow::anyhow!("Failed to resolve server address {}: {}", config.ostp.server_addr, e))?
+        .collect();
+        
+
+    let target_addr = resolved_addrs.first()
+        .ok_or_else(|| anyhow::anyhow!("No IP addresses resolved for {}", config.ostp.server_addr))?;
+        
+    log_to_core_file(&format!("[core] Resolved server address to {}", target_addr));
+    config.ostp.server_addr = target_addr.to_string();
+
+
+    #[cfg(target_os = "linux")]
+    if config.mode == "tun" {
+        println!("\n[ostp] ===========================================================================");
+        println!("[ostp] WARNING: You are starting TUN mode on a Linux system.");
+        println!("[ostp] If this is a remote headless server, routing all traffic through the TUN");
+        println!("[ostp] interface WILL DROP your SSH connection and lock you out!");
+        println!("[ostp] ");
+        println!("[ostp] SOLUTION: Add a static route for your client IP to bypass the TUN.");
+        println!("[ostp] Find your default gateway (ip route | grep default) and run:");
+        println!("[ostp]   sudo ip route add <your-client-ip> via <default-gateway-ip>");
+        println!("[ostp] ===========================================================================\n");
+    }
+
+    #[cfg(target_os = "linux")]
+    if config.mode == "proxy" {
+        println!("\n[ostp] ===========================================================================");
+        println!("[ostp] Proxy mode initialized on {}", config.local_proxy.bind_addr);
+        println!("[ostp] ===========================================================================\n");
+    }
+
+    let _sysproxy_guard = if config.mode == "proxy" {
+        // Enable system proxy and set initial ProxyOverride with user exclusions
+        let guard = Some(crate::sysproxy::SystemProxyGuard::enable(&config.local_proxy.bind_addr));
+        crate::sysproxy::update_proxy_bypass_list(
+            &config.exclusions.domains,
+            &config.exclusions.ips,
+        );
+        guard
+    } else {
+        None
+    };
+
+    let (proxy_events_tx, proxy_events_rx) = mpsc::channel(256);
+    let (client_msgs_tx, client_msgs_rx) = mpsc::unbounded_channel();
+    
+    // Setup exclusions hot-reload channel
+    let (reload_tx, reload_rx) = watch::channel(config.exclusions.clone());
+
+    let mut bridge = Bridge::new(&config, metrics)?;
+    bridge.reload_tx = Some(reload_tx.clone());
 
     let (ui_tx, mut ui_rx) = mpsc::channel(512);
     let (cmd_tx, cmd_rx) = mpsc::channel(128);
     let (shutdown_tx, shutdown_rx) = watch::channel(false);
     let proxy_shutdown_rx = shutdown_tx.subscribe();
 
-    let is_tun = config.mode == "tun";
 
     // Auto-connect on startup
     let _ = cmd_tx.send(BridgeCommand::ToggleTunnel).await;
@@ -147,28 +334,25 @@ pub async fn run_client(config: crate::config::ClientConfig) -> Result<()> {
             match msg {
                 crate::app::UiEvent::Log(text) => {
                     if debug_enabled || is_essential_log(&text) {
-                        println!("[client] {text}");
+                        log_to_core_file(&format!("[ostp] {text}"));
+                        println!("[ostp] {text}");
                     }
                 }
                 crate::app::UiEvent::Metrics { status, rtt_ms, .. } => {
                     let status_str = status.as_str().to_string();
                     if last_status != Some(status_str.clone()) {
                         last_status = Some(status_str.clone());
-                        println!("[client] status={status_str} rtt_ms={:.1}", rtt_ms);
+                        println!("[ostp] Status: {} (rtt={:.1}ms)", status_str, rtt_ms);
                     }
                 }
                 crate::app::UiEvent::Traffic { .. } => {}
                 crate::app::UiEvent::ProfileChanged(profile) => {
                     if debug_enabled {
-                        println!("[client] profile={profile:?}");
+                        println!("[ostp] Obfuscation profile: {profile:?}");
                     }
                 }
                 crate::app::UiEvent::TunnelStopped => {
-                    if is_tun {
-                        println!("[client] tunnel=tun stopped, reconnecting in 5s");
-                    } else {
-                        println!("[client] tunnel=proxy stopped, reconnecting in 5s");
-                    }
+                    println!("[ostp] Connection interrupted. Reconnecting in 5 seconds...");
                     let cmd_tx_inner = cmd_tx_clone.clone();
                     tokio::spawn(async move {
                         tokio::time::sleep(tokio::time::Duration::from_secs(5)).await;
@@ -179,16 +363,17 @@ pub async fn run_client(config: crate::config::ClientConfig) -> Result<()> {
         }
     });
 
-    let bridge_task = tokio::spawn(async move {
+    let mut bridge_task = tokio::spawn(async move {
         bridge.run(ui_tx, cmd_rx, shutdown_rx, proxy_events_rx, client_msgs_tx).await
     });
 
     let config_clone = config.clone();
-    let proxy_task = tokio::spawn(async move {
+    let proxy_exclusions_rx = reload_rx.clone();
+    let mut proxy_task = tokio::spawn(async move {
         tunnel::run_local_proxy(
             config.local_proxy,
             config.ostp,
-            config.exclusions,
+            proxy_exclusions_rx,
             config.debug,
             proxy_shutdown_rx,
             proxy_events_tx,
@@ -198,23 +383,76 @@ pub async fn run_client(config: crate::config::ClientConfig) -> Result<()> {
     });
 
     let wintun_shutdown_rx = shutdown_tx.subscribe();
-    let wintun_task = if config_clone.mode == "tun" {
+    let wintun_exclusions_rx = reload_rx.clone();
+    let mut wintun_task = if config_clone.mode == "tun" {
         Some(tokio::spawn(async move {
-            tunnel::run_tun_tunnel(config_clone, wintun_shutdown_rx).await
+            tunnel::run_tun_tunnel(config_clone, wintun_shutdown_rx, wintun_exclusions_rx).await
         }))
     } else {
         None
     };
 
-    // Wait for Ctrl-C / signal
-    wait_for_shutdown_signal().await?;
-    let _ = cmd_tx.send(BridgeCommand::Shutdown).await;
+    // Wait for local_shutdown
+    let mut local_shutdown = shutdown_rx_ext.clone();
+    let cmd_tx_loop = cmd_tx.clone();
+    tokio::spawn(async move {
+        loop {
+            tokio::select! {
+                _ = local_shutdown.changed() => {
+                    if *local_shutdown.borrow() {
+                        let _ = cmd_tx_loop.send(BridgeCommand::Shutdown).await;
+                        break;
+                    }
+                }
+                Some(Ok(_)) = async {
+                    if let Some(ref mut rx) = config_rx {
+                        Some(rx.changed().await)
+                    } else {
+                        std::future::pending().await
+                    }
+                } => {
+                    if let Some(ref rx) = config_rx {
+                        let new_cfg = rx.borrow().clone();
+                        // Update Windows ProxyOverride so excluded domains/IPs
+                        // bypass the system proxy immediately (proxy mode only).
+                        crate::sysproxy::update_proxy_bypass_list(
+                            &new_cfg.exclusions.domains,
+                            &new_cfg.exclusions.ips,
+                        );
+                        let _ = reload_tx.send(new_cfg.exclusions);
+                    }
+                }
+            }
+        }
+    });
 
-    let _ = shutdown_tx.send(true);
-    let _ = bridge_task.await?;
-    let _ = proxy_task.await?;
+    // Wait for either external shutdown OR any task to fail
+    tokio::select! {
+        _ = shutdown_rx_ext.changed() => {
+            let _ = cmd_tx.send(BridgeCommand::Shutdown).await;
+            let _ = shutdown_tx.send(true);
+        }
+        res = &mut bridge_task => {
+            let _ = shutdown_tx.send(true);
+            res.map_err(|e| anyhow::anyhow!("Bridge task panicked: {}", e))??;
+        }
+        res = &mut proxy_task => {
+            let _ = shutdown_tx.send(true);
+            res.map_err(|e| anyhow::anyhow!("Proxy task panicked: {}", e))??;
+        }
+        res = async {
+            if let Some(t) = wintun_task.as_mut() { t.await } else { std::future::pending().await }
+        } => {
+            let _ = shutdown_tx.send(true);
+            res.map_err(|e| anyhow::anyhow!("TUN task panicked: {}", e))??;
+        }
+    }
+
+    // Final cleanup: wait for tasks to finish
+    let _ = bridge_task.await;
+    let _ = proxy_task.await;
     if let Some(task) = wintun_task {
-        let _ = task.await?;
+        let _ = task.await;
     }
 
     Ok(())
@@ -234,15 +472,17 @@ fn format_bytes(bps: u64) -> String {
 fn is_essential_log(text: &str) -> bool {
     matches!(
         text,
-        "Handshaking started"
-            | "Bridge connection established"
-            | "TUN Tunnel established"
+        "Connection established"
+            | "TUN tunnel established"
+            | "TUN tunnel stopped"
             | "Bridge stopped"
-            | "TUN Tunnel stopped"
             | "Runtime config reloaded"
-    ) || text.starts_with("Connected UDP directly to ")
-        || text.starts_with("TURN: Relay allocated")
+            | "Connecting to remote server..."
+    ) || text.starts_with("Connected to ")
+        || text.starts_with("TURN relay allocated")
         || text.starts_with("TURN allocation failed")
-        || text.starts_with("Handshake failed")
-        || text.starts_with("Connection timeout")
+        || text.starts_with("Allocating TURN relay")
+        || text.starts_with("Connection failed:")
+        || text.starts_with("Connection lost")
+        || text.starts_with("Protocol tick fatal error")
 }

ostp-client/src/signal.rs

@@ -8,8 +8,12 @@ pub async fn wait_for_shutdown_signal() -> Result<()> {
     let mut sigint = signal(SignalKind::interrupt())?;
 
     tokio::select! {
-        _ = sigterm.recv() => {}
-        _ = sigint.recv() => {}
+        _ = sigterm.recv() => {
+            tracing::info!("Received SIGTERM, shutting down");
+        }
+        _ = sigint.recv() => {
+            tracing::info!("Received SIGINT, shutting down");
+        }
     }
 
     Ok(())
@@ -17,6 +21,37 @@ pub async fn wait_for_shutdown_signal() -> Result<()> {
 
 #[cfg(not(unix))]
 pub async fn wait_for_shutdown_signal() -> Result<()> {
-    tokio::signal::ctrl_c().await?;
+    #[cfg(target_os = "windows")]
+    {
+        use tokio::signal::windows::{ctrl_break, ctrl_c, ctrl_close};
+        let mut c_c = ctrl_c()?;
+        let mut c_close = ctrl_close()?;
+        let mut c_break = ctrl_break()?;
+
+        tokio::select! {
+            res = c_c.recv() => {
+                tracing::info!("Received Ctrl+C, shutting down");
+                if res.is_none() {
+                    std::future::pending::<()>().await;
+                }
+            }
+            res = c_close.recv() => {
+                tracing::info!("Received console close event, shutting down");
+                if res.is_none() {
+                    std::future::pending::<()>().await;
+                }
+            }
+            res = c_break.recv() => {
+                tracing::info!("Received Ctrl+Break, shutting down");
+                if res.is_none() {
+                    std::future::pending::<()>().await;
+                }
+            }
+        }
+    }
+    #[cfg(not(target_os = "windows"))]
+    {
+        tokio::signal::ctrl_c().await?;
+    }
     Ok(())
 }

ostp-client/src/sysproxy.rs

@@ -1,6 +1,12 @@
 #[cfg(target_os = "windows")]
 use std::process::Command;
 
+#[cfg(target_os = "windows")]
+use std::os::windows::process::CommandExt;
+
+#[cfg(target_os = "windows")]
+const CREATE_NO_WINDOW: u32 = 0x08000000;
+
 #[cfg(target_os = "windows")]
 #[link(name = "wininet")]
 extern "system" {
@@ -18,50 +24,143 @@ const INTERNET_OPTION_REFRESH: u32 = 37;
 
 #[cfg(target_os = "windows")]
 pub fn enable_windows_proxy(proxy_addr: &str) {
-    let _ = Command::new("reg")
-        .args([
-            "add",
-            "HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Internet Settings",
-            "/v",
-            "ProxyEnable",
-            "/t",
-            "REG_DWORD",
-            "/d",
-            "1",
-            "/f",
-        ])
-        .output();
-        
-    let proxy_str = proxy_addr.to_string();
-    let _ = Command::new("reg")
-        .args([
-            "add",
-            "HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Internet Settings",
-            "/v",
-            "ProxyServer",
-            "/t",
-            "REG_SZ",
-            "/d",
-            &proxy_str,
-            "/f",
-        ])
-        .output();
+    tracing::info!("Enabling Windows system proxy: {}", proxy_addr);
 
+    let result = Command::new("reg")
+        .creation_flags(CREATE_NO_WINDOW)
+        .args([
+            "add",
+            "HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Internet Settings",
+            "/v", "ProxyEnable",
+            "/t", "REG_DWORD",
+            "/d", "1",
+            "/f",
+        ])
+        .output();
+    match result {
+        Ok(out) if !out.status.success() => {
+            tracing::error!("Failed to set ProxyEnable: {}", String::from_utf8_lossy(&out.stderr));
+        }
+        Err(e) => tracing::error!("Failed to execute reg.exe (ProxyEnable): {}", e),
+        _ => {}
+    }
+
+    let result = Command::new("reg")
+        .creation_flags(CREATE_NO_WINDOW)
+        .args([
+            "add",
+            "HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Internet Settings",
+            "/v", "ProxyServer",
+            "/t", "REG_SZ",
+            "/d", proxy_addr,
+            "/f",
+        ])
+        .output();
+    match result {
+        Ok(out) if !out.status.success() => {
+            tracing::error!("Failed to set ProxyServer: {}", String::from_utf8_lossy(&out.stderr));
+        }
+        Err(e) => tracing::error!("Failed to execute reg.exe (ProxyServer): {}", e),
+        _ => {}
+    }
+
+    // Set initial bypass list (will be expanded by update_proxy_bypass_list)
+    update_proxy_bypass_list_windows(&[], &[]);
+
+    refresh_wininet();
+    tracing::info!("System proxy enabled successfully");
+}
+
+/// Update the Windows ProxyOverride registry value to include user-configured
+/// excluded domains and IPs. This makes excluded hosts bypass the OSTP proxy
+/// entirely at the OS level — the most reliable split-tunneling mechanism.
+///
+/// For each domain `d`, adds both `d` and `*.d` so both the root and all
+/// subdomains bypass the proxy.
+/// For IPs, adds them verbatim (Windows supports exact IPs and wildcards like
+/// `192.168.*`).
+#[cfg(target_os = "windows")]
+pub fn update_proxy_bypass_list(domains: &[String], ips: &[String]) {
+    update_proxy_bypass_list_windows(domains, ips);
     refresh_wininet();
 }
 
+#[cfg(not(target_os = "windows"))]
+pub fn update_proxy_bypass_list(_domains: &[String], _ips: &[String]) {
+    // Linux/macOS: no-op (gnome/kde proxy bypass list update not implemented)
+}
+
 #[cfg(target_os = "windows")]
-pub fn disable_windows_proxy() {
+fn update_proxy_bypass_list_windows(domains: &[String], ips: &[String]) {
+    // Base list: always bypass local addresses
+    let mut parts: Vec<String> = vec![
+        "localhost".into(),
+        "127.*".into(),
+        "10.*".into(),
+        "172.16.*".into(),
+        "172.17.*".into(),
+        "172.18.*".into(),
+        "172.19.*".into(),
+        "172.20.*".into(),
+        "172.21.*".into(),
+        "172.22.*".into(),
+        "172.23.*".into(),
+        "172.24.*".into(),
+        "172.25.*".into(),
+        "172.26.*".into(),
+        "172.27.*".into(),
+        "172.28.*".into(),
+        "172.29.*".into(),
+        "172.30.*".into(),
+        "172.31.*".into(),
+        "192.168.*".into(),
+        "<local>".into(),
+    ];
+
+    // Add excluded domains: both exact and wildcard subdomain form
+    for d in domains {
+        let d = d.trim().trim_start_matches('.').to_lowercase();
+        if d.is_empty() { continue; }
+        parts.push(d.clone());
+        parts.push(format!("*.{}", d));
+    }
+
+    // Add excluded IPs verbatim
+    for ip in ips {
+        let ip = ip.trim();
+        if ip.is_empty() { continue; }
+        // Strip CIDR suffix if present — Windows ProxyOverride doesn't support CIDR
+        let host = ip.split('/').next().unwrap_or(ip);
+        parts.push(host.to_string());
+    }
+
+    let override_value = parts.join(";");
+    tracing::info!("Updating ProxyOverride: {}", override_value);
+
     let _ = Command::new("reg")
+        .creation_flags(CREATE_NO_WINDOW)
         .args([
             "add",
             "HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Internet Settings",
-            "/v",
-            "ProxyEnable",
-            "/t",
-            "REG_DWORD",
-            "/d",
-            "0",
+            "/v", "ProxyOverride",
+            "/t", "REG_SZ",
+            "/d", &override_value,
+            "/f",
+        ])
+        .output();
+}
+
+#[cfg(target_os = "windows")]
+pub fn disable_system_proxy() {
+    tracing::info!("Disabling Windows system proxy");
+    let _ = Command::new("reg")
+        .creation_flags(CREATE_NO_WINDOW)
+        .args([
+            "add",
+            "HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Internet Settings",
+            "/v", "ProxyEnable",
+            "/t", "REG_DWORD",
+            "/d", "0",
             "/f",
         ])
         .output();
@@ -88,26 +187,92 @@ fn refresh_wininet() {
 }
 
 #[cfg(not(target_os = "windows"))]
-pub fn enable_windows_proxy(_proxy_addr: &str) {}
+pub fn enable_system_proxy(proxy_addr: &str) {
+    let parts: Vec<&str> = proxy_addr.split(':').collect();
+    let host = parts.first().unwrap_or(&"127.0.0.1");
+    let port = parts.get(1).unwrap_or(&"1088");
+
+    let is_gui = std::env::var("DISPLAY").is_ok() || std::env::var("WAYLAND_DISPLAY").is_ok();
+
+    if is_gui {
+        tracing::info!("Enabling Linux system proxy (GNOME/KDE): {}", proxy_addr);
+        
+        // Try GNOME gsettings
+        let gnome_res = std::process::Command::new("gsettings")
+            .args(["set", "org.gnome.system.proxy", "mode", "manual"])
+            .output();
+            
+        if let Ok(out) = gnome_res {
+            if out.status.success() {
+                let _ = std::process::Command::new("gsettings").args(["set", "org.gnome.system.proxy.socks", "host", host]).output();
+                let _ = std::process::Command::new("gsettings").args(["set", "org.gnome.system.proxy.socks", "port", port]).output();
+                let _ = std::process::Command::new("gsettings").args(["set", "org.gnome.system.proxy", "ignore-hosts", "['localhost', '127.0.0.0/8', '10.0.0.0/8', '192.168.0.0/16']"]).output();
+                tracing::info!("GNOME system proxy enabled.");
+                return;
+            }
+        }
+
+        // Try KDE kwriteconfig5/6
+        for cmd in ["kwriteconfig5", "kwriteconfig6"] {
+            let kde_res = std::process::Command::new(cmd)
+                .args(["--file", "kioslaverc", "--group", "Proxy Settings", "--key", "ProxyType", "1"])
+                .output();
+                
+            if let Ok(out) = kde_res {
+                if out.status.success() {
+                    let socks_val = format!("socks://{}:{}", host, port);
+                    let _ = std::process::Command::new(cmd).args(["--file", "kioslaverc", "--group", "Proxy Settings", "--key", "socksProxy", &socks_val]).output();
+                    let _ = std::process::Command::new("dbus-send").args(["--type=signal", "/KIO/Scheduler", "org.kde.KIO.Scheduler.reparseSlaveConfiguration", "string:''"]).output();
+                    tracing::info!("KDE system proxy enabled.");
+                    return;
+                }
+            }
+        }
+    }
+
+    // Headless fallback
+    println!("\n===================================================================");
+    println!("OSTP Local Proxy is running at socks5://{}", proxy_addr);
+    println!("Since you are in a headless/terminal environment, OSTP cannot automatically");
+    println!("configure your system proxy. To route traffic from this terminal, run:");
+    println!("\n    eval $(ostp proxy-env)\n");
+    println!("Or configure your application (e.g. curl -x socks5://{})", proxy_addr);
+    println!("===================================================================\n");
+}
 
 #[cfg(not(target_os = "windows"))]
-pub fn disable_windows_proxy() {}
+pub fn disable_system_proxy() {
+    let is_gui = std::env::var("DISPLAY").is_ok() || std::env::var("WAYLAND_DISPLAY").is_ok();
+    if is_gui {
+        tracing::info!("Disabling Linux system proxy...");
+        let _ = std::process::Command::new("gsettings").args(["set", "org.gnome.system.proxy", "mode", "none"]).output();
+        let _ = std::process::Command::new("kwriteconfig5").args(["--file", "kioslaverc", "--group", "Proxy Settings", "--key", "ProxyType", "0"]).output();
+        let _ = std::process::Command::new("kwriteconfig6").args(["--file", "kioslaverc", "--group", "Proxy Settings", "--key", "ProxyType", "0"]).output();
+        let _ = std::process::Command::new("dbus-send").args(["--type=signal", "/KIO/Scheduler", "org.kde.KIO.Scheduler.reparseSlaveConfiguration", "string:''"]).output();
+    }
+}
 
-pub struct WindowsProxyGuard {
+#[cfg(target_os = "windows")]
+pub fn enable_system_proxy(proxy_addr: &str) {
+    enable_windows_proxy(proxy_addr);
+}
+
+
+pub struct SystemProxyGuard {
     active: bool,
 }
 
-impl WindowsProxyGuard {
+impl SystemProxyGuard {
     pub fn enable(proxy_addr: &str) -> Self {
-        enable_windows_proxy(proxy_addr);
+        enable_system_proxy(proxy_addr);
         Self { active: true }
     }
 }
 
-impl Drop for WindowsProxyGuard {
+impl Drop for SystemProxyGuard {
     fn drop(&mut self) {
         if self.active {
-            disable_windows_proxy();
+            disable_system_proxy();
         }
     }
 }

ostp-client/src/transport/mod.rs

@@ -0,0 +1,56 @@
+
+use std::sync::Arc;
+use tokio::net::UdpSocket;
+use bytes::Bytes;
+
+#[derive(Clone)]
+pub enum Transport {
+    Udp(Arc<UdpSocket>),
+    Uot {
+        tx: tokio::sync::mpsc::Sender<Bytes>,
+        rx: Arc<tokio::sync::Mutex<tokio::sync::mpsc::Receiver<Bytes>>>,
+    }
+}
+
+impl Transport {
+    pub async fn send(&self, frame: &Bytes) -> std::io::Result<usize> {
+        match self {
+            Self::Udp(sock) => sock.send(frame).await,
+            Self::Uot { tx, .. } => {
+                tx.send(frame.clone()).await.map_err(|_| std::io::Error::new(std::io::ErrorKind::BrokenPipe, "uot closed"))?;
+                Ok(frame.len())
+            }
+        }
+    }
+
+    pub async fn send_to(&self, frame: &Bytes, target: std::net::SocketAddr) -> std::io::Result<usize> {
+        match self {
+            Self::Udp(sock) => sock.send_to(frame, target).await,
+            Self::Uot { .. } => self.send(frame).await,
+        }
+    }
+
+    pub async fn recv(&self, buf: &mut [u8]) -> std::io::Result<usize> {
+        match self {
+            Self::Udp(sock) => sock.recv(buf).await,
+            Self::Uot { rx, .. } => {
+                let mut rx = rx.lock().await;
+                match rx.recv().await {
+                    Some(bytes) => {
+                        let len = bytes.len().min(buf.len());
+                        buf[..len].copy_from_slice(&bytes[..len]);
+                        Ok(len)
+                    }
+                    None => Err(std::io::Error::new(std::io::ErrorKind::BrokenPipe, "uot closed")),
+                }
+            }
+        }
+    }
+
+    pub fn local_addr(&self) -> std::io::Result<std::net::SocketAddr> {
+        match self {
+            Self::Udp(sock) => sock.local_addr(),
+            Self::Uot { .. } => Ok("0.0.0.0:0".parse().unwrap()),
+        }
+    }
+}

ostp-client/src/tunnel/exclusion.rs

@@ -0,0 +1,134 @@
+use crate::config::ExclusionConfig;
+use std::time::Duration;
+use tokio::time::timeout;
+
+#[derive(Clone)]
+pub struct ExclusionMatcher {
+    pub domain_suffix: Vec<String>,
+    pub cidrs: Vec<Cidr>,
+    pub processes: Vec<String>,
+    pub physical_if_index: Option<u32>,
+    pub physical_if_name: Option<String>,
+}
+
+impl ExclusionMatcher {
+    pub fn new(
+        exclusions: &ExclusionConfig,
+        physical_if_index: Option<u32>,
+        physical_if_name: Option<String>,
+    ) -> Self {
+        let mut cidrs = Vec::new();
+        for ip in &exclusions.ips {
+            if let Some(cidr) = parse_cidr(ip) {
+                cidrs.push(cidr);
+            }
+        }
+
+        let processes = exclusions.processes.iter()
+            .map(|p| p.trim().to_lowercase())
+            .filter(|p| !p.is_empty())
+            .collect();
+
+        Self {
+            domain_suffix: exclusions
+                .domains
+                .iter()
+                .map(|d| d.trim().trim_start_matches('.').to_lowercase())
+                .filter(|d| !d.is_empty())
+                .collect(),
+            cidrs,
+            processes,
+            physical_if_index,
+            physical_if_name,
+        }
+    }
+
+    pub async fn should_bypass_target(&self, host: &str, port: u16, timeout_value: Duration) -> bool {
+        if self.match_domain(host) {
+            return true;
+        }
+
+        if self.cidrs.is_empty() {
+            return false;
+        }
+
+        if let Ok(ip) = host.parse::<std::net::IpAddr>() {
+            return self.match_ip(&ip);
+        }
+
+        let lookup_target = (host.to_string(), port);
+        match timeout(timeout_value, tokio::net::lookup_host(lookup_target)).await {
+            Ok(Ok(addrs)) => addrs.into_iter().any(|addr| self.match_ip(&addr.ip())),
+            _ => false,
+        }
+    }
+
+    pub fn match_domain(&self, host: &str) -> bool {
+        if self.domain_suffix.is_empty() {
+            return false;
+        }
+        let host = host.trim_end_matches('.').to_lowercase();
+        self.domain_suffix.iter().any(|suffix| {
+            host == *suffix || host.ends_with(&format!(".{suffix}"))
+        })
+    }
+
+    pub fn match_ip(&self, ip: &std::net::IpAddr) -> bool {
+        self.cidrs.iter().any(|cidr| cidr.contains(ip))
+    }
+
+    pub fn match_process(&self, process_name: &str) -> bool {
+        if self.processes.is_empty() {
+            return false;
+        }
+        let p = process_name.to_lowercase();
+        self.processes.iter().any(|ex| p.contains(ex))
+    }
+}
+
+#[derive(Clone)]
+pub enum Cidr {
+    V4(u32, u8),
+    V6(u128, u8),
+}
+
+impl Cidr {
+    pub fn contains(&self, ip: &std::net::IpAddr) -> bool {
+        match (self, ip) {
+            (Cidr::V4(net, bits), std::net::IpAddr::V4(addr)) => {
+                let mask = if *bits == 0 { 0 } else { u32::MAX << (32 - bits) };
+                let ip = u32::from_be_bytes(addr.octets());
+                (ip & mask) == (*net & mask)
+            }
+            (Cidr::V6(net, bits), std::net::IpAddr::V6(addr)) => {
+                let mask = if *bits == 0 { 0 } else { u128::MAX << (128 - bits) };
+                let ip = u128::from_be_bytes(addr.octets());
+                (ip & mask) == (*net & mask)
+            }
+            _ => false,
+        }
+    }
+}
+
+pub fn parse_cidr(s: &str) -> Option<Cidr> {
+    let parts: Vec<&str> = s.split('/').collect();
+    if parts.is_empty() || parts.len() > 2 {
+        return None;
+    }
+    if let Ok(ip) = parts[0].parse::<std::net::IpAddr>() {
+        let bits = if parts.len() == 2 {
+            parts[1].parse::<u8>().ok()?
+        } else {
+            match ip {
+                std::net::IpAddr::V4(_) => 32,
+                std::net::IpAddr::V6(_) => 128,
+            }
+        };
+        match ip {
+            std::net::IpAddr::V4(v4) => Some(Cidr::V4(u32::from_be_bytes(v4.octets()), bits)),
+            std::net::IpAddr::V6(v6) => Some(Cidr::V6(u128::from_be_bytes(v6.octets()), bits)),
+        }
+    } else {
+        None
+    }
+}

ostp-client/src/tunnel/linux_handler.rs

@@ -1,205 +0,0 @@
-use anyhow::{anyhow, Result};
-use tokio::sync::watch;
-
-#[cfg(target_os = "linux")]
-use std::net::ToSocketAddrs;
-#[cfg(target_os = "linux")]
-use std::process::{Command, Stdio, Child};
-#[cfg(target_os = "linux")]
-use std::io::{BufRead, BufReader};
-
-#[cfg(target_os = "linux")]
-struct LinuxRouteGuard {
-    server_ip_str: String,
-    default_gw: String,
-    default_if: String,
-    child: Option<Child>,
-}
-
-#[cfg(target_os = "linux")]
-impl Drop for LinuxRouteGuard {
-    fn drop(&mut self) {
-        if let Some(mut child) = self.child.take() {
-            let _ = child.kill();
-        }
-        let cleanup_script = format!(
-            "ip route del 0.0.0.0/1 dev ostp_tun || true; \
-             ip route del 128.0.0.0/1 dev ostp_tun || true; \
-             ip route del {} via {} dev {} || true; \
-             ip route del 1.1.1.1 via {} dev {} || true; \
-             ip link set dev ostp_tun down || true; \
-             ip tuntap del name ostp_tun mode tun || true",
-            self.server_ip_str, self.default_gw, self.default_if,
-            self.default_gw, self.default_if
-        );
-        let _ = Command::new("sh").args(["-c", &cleanup_script]).output();
-    }
-}
-
-#[cfg(target_os = "linux")]
-pub async fn run_linux_tunnel(
-    config: crate::config::ClientConfig,
-    mut shutdown: watch::Receiver<bool>,
-) -> Result<()> {
-    let debug = config.debug;
-    if debug {
-        println!("[ostp-client] Starting Linux TUN handler initialization...");
-    }
-
-    let exe = std::env::current_exe()?;
-    let dir = exe.parent().ok_or_else(|| anyhow!("failed to get binary directory"))?;
-    
-    let mut tun2socks_exe = dir.join("tun2socks");
-    if !tun2socks_exe.exists() {
-        // Try system PATH via standard command check
-        let in_path = Command::new("which")
-            .arg("tun2socks")
-            .output()
-            .map(|o| o.status.success())
-            .unwrap_or(false);
-
-        if in_path {
-            tun2socks_exe = std::path::PathBuf::from("tun2socks");
-        } else {
-            return Err(anyhow!("tun2socks executable not found in local dir or PATH. Please ensure dependencies are present."));
-        }
-    }
-
-    // 2. Resolve Server IP for routing table exclusion
-    let server_ip = config.ostp.server_addr.to_socket_addrs()
-        .map_err(|e| anyhow!("Failed to resolve remote server IP: {}", e))?
-        .next()
-        .map(|addr| addr.ip())
-        .ok_or_else(|| anyhow!("Could not resolve host IP for routing exclusion"))?;
-    
-    let server_ip_str = server_ip.to_string();
-
-    if debug {
-        println!("[ostp-client] Resolved remote server IP: {}", server_ip_str);
-    }
-
-    // 3. Detect current default gateway and interface
-    let route_output = Command::new("sh")
-        .arg("-c")
-        .arg("ip route show default | head -n1")
-        .output()?;
-    
-    let route_str = String::from_utf8_lossy(&route_output.stdout);
-    let parts: Vec<&str> = route_str.split_whitespace().collect();
-    
-    // Expected: "default via 192.168.1.1 dev eth0 ..."
-    let mut default_gw = String::new();
-    let mut default_if = String::new();
-    
-    for i in 0..parts.len() {
-        if parts[i] == "via" && i + 1 < parts.len() {
-            default_gw = parts[i+1].to_string();
-        }
-        if parts[i] == "dev" && i + 1 < parts.len() {
-            default_if = parts[i+1].to_string();
-        }
-    }
-
-    if default_gw.is_empty() || default_if.is_empty() {
-        return Err(anyhow!("Failed to discover active default gateway or network interface on Linux system."));
-    }
-
-    if debug {
-        println!("[ostp-client] Physical route anchor: gateway={} interface={}", default_gw, default_if);
-    }
-
-    // 4. Setup commands (Using standard /1 routing trick for fail-proof overriding)
-    let setup_script = format!(
-        "ip tuntap add name ostp_tun mode tun || true; \
-         ip addr add 10.1.0.2/24 dev ostp_tun || true; \
-         ip link set dev ostp_tun up; \
-         ip route add {} via {} dev {}; \
-         ip route add 1.1.1.1 via {} dev {}; \
-         ip route add 0.0.0.0/1 dev ostp_tun; \
-         ip route add 128.0.0.0/1 dev ostp_tun",
-        server_ip_str, default_gw, default_if,
-        default_gw, default_if
-    );
-
-    if debug {
-        println!("[ostp-client] Executing Linux network config: {}", setup_script);
-    }
-
-    let out = Command::new("sh")
-        .args(["-c", &setup_script])
-        .output()?;
-    
-    if !out.status.success() && debug {
-        println!("[ostp-client] Warning: Setup routing returned: {}", String::from_utf8_lossy(&out.stderr));
-    }
-
-    // 5. Prepare and launch tun2socks
-    // Using HTTP Proxy natively avoids any UDP Associate requests,
-    // providing clean TCP proxying with maximum reliability.
-    let proxy_url = format!("http://{}", config.local_proxy.bind_addr);
-    
-    if debug {
-        println!("[ostp-client] Spawning {} -device ostp_tun -proxy {}", tun2socks_exe.display(), proxy_url);
-    }
-
-    let mut child = Command::new(&tun2socks_exe)
-        .args([
-            "-device", "ostp_tun",
-            "-proxy", &proxy_url,
-        ])
-        .stdout(if debug { Stdio::piped() } else { Stdio::null() })
-        .stderr(if debug { Stdio::piped() } else { Stdio::null() })
-        .spawn()
-        .map_err(|e| anyhow!("Failed to spawn tun2socks process: {}", e))?;
-
-    let mut _guard = LinuxRouteGuard {
-        server_ip_str: server_ip_str.clone(),
-        default_gw: default_gw.clone(),
-        default_if: default_if.clone(),
-        child: None,
-    };
-
-    println!("[client] TUN Tunnel established, Linux traffic is now routing through OSTP.");
-
-    if debug {
-        let stdout = child.stdout.take().unwrap();
-        let stderr = child.stderr.take().unwrap();
-        
-        tokio::spawn(async move {
-            let reader = BufReader::new(stdout);
-            for line in reader.lines().map_while(Result::ok) {
-                println!("[tun2socks] {}", line);
-            }
-        });
-        
-        tokio::spawn(async move {
-            let reader = BufReader::new(stderr);
-            for line in reader.lines().map_while(Result::ok) {
-                eprintln!("[tun2socks-err] {}", line);
-            }
-        });
-    }
-
-    _guard.child = Some(child);
-
-    // 6. Wait for shutdown signal
-    let _ = shutdown.changed().await;
-
-    println!("[client] Deactivating TUN tunnel and restoring Linux network topology...");
-
-    // Drop guard runs cleanup automatically
-    drop(_guard);
-
-    println!("[client] Linux TUN Tunnel stopped.");
-    
-    Ok(())
-}
-
-#[cfg(not(target_os = "linux"))]
-#[allow(dead_code)]
-pub async fn run_linux_tunnel(
-    _config: crate::config::ClientConfig,
-    _shutdown: watch::Receiver<bool>,
-) -> Result<()> {
-    Err(anyhow!("Linux tunnel driver executed on a non-Linux host!"))
-}

ostp-client/src/tunnel/mod.rs

@@ -1,27 +1,14 @@
 mod proxy;
-mod wintun_handler;
-mod linux_handler;
+pub mod native_handler;
+
+mod udp_nat;
 
 pub async fn run_tun_tunnel(
     config: crate::config::ClientConfig,
-    shutdown: watch::Receiver<bool>,
+    shutdown: tokio::sync::watch::Receiver<bool>,
+    exclusions_rx: tokio::sync::watch::Receiver<crate::config::ExclusionConfig>,
 ) -> anyhow::Result<()> {
-    #[cfg(target_os = "windows")]
-    {
-        wintun_handler::run_wintun_tunnel(config, shutdown).await
-    }
-
-    #[cfg(target_os = "linux")]
-    {
-        linux_handler::run_linux_tunnel(config, shutdown).await
-    }
-
-    #[cfg(not(any(target_os = "windows", target_os = "linux")))]
-    {
-        let _ = shutdown;
-        let _ = config;
-        anyhow::bail!("Operating system unsupported, text an issue at github.");
-    }
+    native_handler::run_native_tunnel(config, shutdown, exclusions_rx).await
 }
 
 use tokio::sync::{mpsc, watch};
@@ -36,6 +23,14 @@ pub enum ProxyEvent {
         stream_id: u16,
         target: String,
     },
+    UdpAssociate {
+        stream_id: u16,
+    },
+    UdpData {
+        stream_id: u16,
+        target: String,
+        payload: bytes::Bytes,
+    },
     Data {
         stream_id: u16,
         payload: bytes::Bytes,
@@ -49,6 +44,7 @@ pub enum ProxyEvent {
 pub enum ProxyToClientMsg {
     ConnectOk,
     Data(bytes::Bytes),
+    UdpData(String, bytes::Bytes),
     Close,
     Error(String),
 }
@@ -57,13 +53,15 @@ pub enum ProxyToClientMsg {
 pub async fn run_local_proxy(
     cfg: LocalProxyConfig,
     ostp: OstpConfig,
-    exclusions: ExclusionConfig,
+    exclusions_rx: watch::Receiver<ExclusionConfig>,
     debug: bool,
     shutdown: watch::Receiver<bool>,
     proxy_events_tx: mpsc::Sender<ProxyEvent>,
-    client_msgs_rx: mpsc::Receiver<(u16, ProxyToClientMsg)>,
+    client_msgs_rx: mpsc::UnboundedReceiver<(u16, ProxyToClientMsg)>,
 ) -> anyhow::Result<()> {
-    run_local_socks5_proxy(cfg, ostp, exclusions, debug, shutdown, proxy_events_tx, client_msgs_rx).await
+    run_local_socks5_proxy(cfg, ostp, exclusions_rx, debug, shutdown, proxy_events_tx, client_msgs_rx).await
 }
 
-
+pub mod exclusion;
+pub mod process_lookup;
+pub mod sni_sniff;

ostp-client/src/tunnel/native_handler.rs

@@ -0,0 +1,744 @@
+use anyhow::{anyhow, Result};
+use tokio::sync::watch;
+
+// ──────────────────────────────────────────────────────────────────────────────
+// Windows / Linux desktop TUN
+// ──────────────────────────────────────────────────────────────────────────────
+
+#[cfg(any(target_os = "windows", target_os = "linux"))]
+pub async fn run_native_tunnel(
+    config: crate::config::ClientConfig,
+    mut shutdown: watch::Receiver<bool>,
+    mut exclusions_rx: watch::Receiver<crate::config::ExclusionConfig>,
+) -> Result<()> {
+    use std::net::ToSocketAddrs;
+    use netstack_smoltcp::StackBuilder;
+    use tokio::io::{AsyncReadExt, AsyncWriteExt};
+    use futures::{StreamExt, SinkExt};
+
+    #[cfg(target_os = "linux")]
+    {
+        use std::io::{self, IsTerminal, Write};
+        if io::stdout().is_terminal() {
+            println!("\n===================================================================");
+            println!("WARNING: TUN mode will modify the system routing table.");
+            println!("If you are connected to a headless server via SSH, you may lose");
+            println!("your connection when default routes are redirected into the tunnel.");
+            println!("===================================================================\n");
+            print!("Are you sure you want to initialize the TUN interface? [yes/no]: ");
+            io::stdout().flush().unwrap();
+
+            let mut input = String::new();
+            io::stdin().read_line(&mut input).unwrap();
+            let ans = input.trim().to_lowercase();
+            if ans != "y" && ans != "yes" {
+                return Err(anyhow!("TUN initialization aborted by user."));
+            }
+        }
+    }
+
+    let debug = config.debug;
+    tracing::info!("Initializing NATIVE TUN tunnel (smoltcp)...");
+
+    // Capture physical interface index for bypass BEFORE we create the TUN device and alter routes.
+    #[cfg(target_os = "windows")]
+    let phys_if_for_bypass: Option<u32> = ostp_tun::windows::windows_route::sys::get_default_ipv4_route().map(|(_, idx)| idx);
+    #[cfg(not(target_os = "windows"))]
+    let phys_if_for_bypass: Option<u32> = None;
+
+    // ── 1. Resolve server IP ──────────────────────────────────────────────────
+    let server_ip = config
+        .ostp
+        .server_addr
+        .to_socket_addrs()
+        .map_err(|e| anyhow!("Failed to resolve server IP: {}", e))?
+        .next()
+        .map(|a| a.ip())
+        .ok_or_else(|| anyhow!("Could not resolve server host"))?;
+    #[allow(unused_variables)]
+    let server_ip_str = server_ip.to_string();
+
+    // ── 2. Resolve excluded domains → IP addresses for bypass routing ─────────
+    let mut bypass_ips: Vec<std::net::IpAddr> = Vec::new();
+
+    // Server IP always bypasses TUN
+    bypass_ips.push(server_ip);
+
+    for ip_str in &config.exclusions.ips {
+        let host = ip_str.split('/').next().unwrap_or(ip_str);
+        if let Ok(ip) = host.parse() {
+            bypass_ips.push(ip);
+        }
+    }
+
+    for domain in &config.exclusions.domains {
+        match tokio::net::lookup_host((domain.as_str(), 443u16)).await {
+            Ok(addrs) => {
+                for addr in addrs {
+                    bypass_ips.push(addr.ip());
+                }
+            }
+            Err(e) => {
+                tracing::warn!("Failed to pre-resolve excluded domain {domain}: {e}");
+            }
+        }
+    }
+
+
+    // ── 3. Create TUN device via ostp-tun crate ───────────────────────────────
+    let opts = ostp_tun::OstpTunOptions {
+        server_ip,
+        bypass_ips,
+        dns_server: config.dns_server.clone(),
+        kill_switch: config.kill_switch,
+        mtu: config.ostp.mtu as u16,
+        wintun_path: None,
+    };
+
+    let tun_interface = ostp_tun::OstpTunInterface::create(opts)
+        .await
+        .map_err(|e| anyhow!("Failed to create OstpTunInterface: {}", e))?;
+        
+    let dev = tun_interface.device;
+    let _route_guard = tun_interface.guard;
+
+    // ── 7. Build smoltcp network stack ────────────────────────────────────────
+    let (stack, tcp_runner, udp_socket, tcp_listener) = StackBuilder::default()
+        .stack_buffer_size(1024)
+        .tcp_buffer_size(1024)
+        .udp_buffer_size(1024)
+        .enable_tcp(true)
+        .enable_udp(true)
+        .mtu(config.ostp.mtu)
+        .build()?;
+
+    let mut runner_task = tokio::spawn(async move {
+        if let Some(runner) = tcp_runner {
+            let _ = runner.await;
+        }
+    });
+
+    // ── 8. Wire TUN ↔ smoltcp stack ───────────────────────────────────────────
+    let (mut stack_sink, mut stack_stream) = stack.split();
+    let (mut tun_read, mut tun_write) = tokio::io::split(dev);
+
+    let mut tun_to_stack = tokio::spawn(async move {
+        let mut buf = vec![0u8; 65536];
+        loop {
+            match tun_read.read(&mut buf).await {
+                Ok(0) => break,
+                Ok(n) => {
+                    let frame = buf[..n].to_vec();
+                    if let Err(e) = stack_sink.send(frame).await {
+                        if e.kind() == std::io::ErrorKind::BrokenPipe {
+                            break;
+                        }
+                    }
+                }
+                Err(e) => {
+                    tracing::debug!("tun_read error: {e}");
+                }
+            }
+        }
+    });
+
+    let mut stack_to_tun = tokio::spawn(async move {
+        while let Some(Ok(frame)) = stack_stream.next().await {
+            if let Err(e) = tun_write.write(&frame).await {
+                tracing::debug!("tun_write error: {e}");
+            }
+        }
+    });
+
+    // ── 9. UDP: forward everything through OSTP proxy ─────────────────────────
+    //   UDP exclusions are handled at the routing table level (step 5), so
+    //   UDP packets for excluded IPs never reach smoltcp at all.
+    let udp_proxy_addr = {
+        let mut a = config.local_proxy.bind_addr.clone();
+        if a.starts_with("0.0.0.0:") {
+            a = a.replace("0.0.0.0:", "127.0.0.1:");
+        }
+        a
+    };
+    // Build exclusion matcher for dynamic bypass
+    let current_exclusions = exclusions_rx.borrow().clone();
+    let matcher = crate::tunnel::exclusion::ExclusionMatcher::new(&current_exclusions, None, None);
+    let matcher_arc = std::sync::Arc::new(tokio::sync::RwLock::new(matcher));
+    
+    let matcher_clone = matcher_arc.clone();
+    tokio::spawn(async move {
+        while let Ok(_) = exclusions_rx.changed().await {
+            let current = exclusions_rx.borrow().clone();
+            let new_matcher = crate::tunnel::exclusion::ExclusionMatcher::new(&current, None, None);
+            *matcher_clone.write().await = new_matcher;
+            if true {
+                tracing::debug!("Desktop TUN exclusions hot-reloaded");
+            }
+        }
+    });
+
+    // Linux: physical interface name for SO_BINDTODEVICE
+    #[cfg(target_os = "linux")]
+    let linux_phys_name = crate::tunnel::proxy::get_linux_physical_if_name();
+    #[cfg(not(target_os = "linux"))]
+    let linux_phys_name: Option<String> = None;
+    let _ = &linux_phys_name; // suppress unused warning on Windows
+
+    let debug_udp = debug;
+    let udp_matcher = matcher_arc.clone();
+    #[cfg(target_os = "linux")]
+    let udp_lin_name = linux_phys_name.clone();
+
+    let mut udp_proxy_task = tokio::spawn(async move {
+        if let Some(udp_sock) = udp_socket {
+            #[cfg(target_os = "linux")]
+            super::udp_nat::run_udp_nat(udp_sock, udp_proxy_addr, debug_udp, udp_matcher, phys_if_for_bypass, udp_lin_name).await;
+            #[cfg(not(target_os = "linux"))]
+            super::udp_nat::run_udp_nat(udp_sock, udp_proxy_addr, debug_udp, udp_matcher, phys_if_for_bypass, None).await;
+        }
+    });
+
+    // ── 10. TCP: forward to OSTP proxy (with domain-level bypass via SNI) ─────
+    //
+    //   For IP-based exclusions:  handled by routing table → packets never arrive here.
+    //   For domain-based exclusions: The IP is already in routing table (pre-resolved in
+    //     step 3), so most traffic won't arrive. As a belt-and-suspenders fallback,
+    //     we also sniff TLS SNI and bypass if it matches — this covers CDN cases where
+    //     the IP wasn't known at startup.
+    //
+    //   For bypassed connections we bind the outgoing socket to the physical interface
+    //   (IP_UNICAST_IF) so it goes out via the real NIC, not TUN.
+
+    let proxy_addr_tcp = {
+        let mut a = config.local_proxy.bind_addr.clone();
+        if a.starts_with("0.0.0.0:") {
+            a = a.replace("0.0.0.0:", "127.0.0.1:");
+        }
+        a
+    };
+
+    // Physical interface index was captured at the start of the function.
+
+    let mut tcp_accept_task = tokio::spawn(async move {
+        let Some(mut listener) = tcp_listener else { return; };
+
+        while let Some((mut stream, local, remote)) = listener.next().await {
+            let proxy_addr = proxy_addr_tcp.clone();
+            let matcher_arc = matcher_arc.clone();
+            #[cfg(target_os = "linux")]
+            let lin_name = linux_phys_name.clone();
+
+            tokio::spawn(async move {
+                let matcher = matcher_arc.read().await.clone();
+                if debug {
+                    tracing::debug!("TUN TCP {local} → {remote}");
+                }
+
+                // ── Sniff TLS ClientHello for SNI ─────────────────────────────
+                let mut sniff_buf = [0u8; 2048];
+                let sniff_len =
+                    match tokio::time::timeout(
+                        std::time::Duration::from_millis(100),
+                        stream.read(&mut sniff_buf),
+                    )
+                    .await
+                    {
+                        Ok(Ok(n)) => n,
+                        _ => 0,
+                    };
+
+                // ── Decide: bypass or tunnel? ─────────────────────────────────
+                let mut should_bypass = false;
+
+                // 1. Process match via OS Extended TCP Table (Windows)
+                #[cfg(target_os = "windows")]
+                if !should_bypass {
+                    if let Some(proc_name) = crate::tunnel::process_lookup::get_process_name_from_port(local.port()) {
+                        if debug {
+                            tracing::debug!("TUN TCP lookup: port {} -> process {}", local.port(), proc_name);
+                        }
+                        if matcher.match_process(&proc_name) {
+                            if debug {
+                                tracing::debug!("TUN TCP BYPASS (Process match): {} → {remote}", proc_name);
+                            }
+                            should_bypass = true;
+                        }
+                    } else {
+                        if debug {
+                            tracing::debug!("TUN TCP lookup: port {} -> no process found", local.port());
+                        }
+                    }
+                }
+
+                // 2. SNI domain check (belt-and-suspenders for CDNs / late-resolved IPs)
+                if !should_bypass && sniff_len > 0 {
+                    if let Some(sni) =
+                        crate::tunnel::sni_sniff::extract_sni(&sniff_buf[..sniff_len])
+                    {
+                        if debug {
+                            tracing::debug!("TUN SNI: {sni}");
+                        }
+                        if matcher.match_domain(&sni) {
+                            if debug {
+                                tracing::info!("TUN TCP BYPASS (SNI domain): {sni} → {remote}");
+                            }
+                            should_bypass = true;
+                        }
+                    }
+                }
+
+                // 3. Destination IP CIDR check (for IPs not in routing table / IPv6)
+                if !should_bypass && matcher.match_ip(&remote.ip()) {
+                    if debug {
+                        tracing::info!("TUN TCP BYPASS (IP match): {remote}");
+                    }
+                    should_bypass = true;
+                }
+
+                // ── Bypass path: direct TCP bypassing TUN ─────────────────────
+                if should_bypass {
+                    let socket = match remote {
+                        std::net::SocketAddr::V4(_) => tokio::net::TcpSocket::new_v4(),
+                        std::net::SocketAddr::V6(_) => tokio::net::TcpSocket::new_v6(),
+                    };
+                    let Ok(socket) = socket else { return; };
+
+                    // Bind to physical interface so packets don't loop back into TUN
+
+                    #[cfg(target_os = "windows")]
+                    if let Some(idx) = phys_if_for_bypass {
+                        if let Err(e) = crate::tunnel::proxy::bind_socket_to_interface(
+                            &socket,
+                            remote.is_ipv6(),
+                            idx,
+                        ) {
+                            tracing::error!("TUN TCP BYPASS failed to bind to physical interface {}: {}", idx, e);
+                        } else {
+                            if debug {
+                                tracing::info!("TUN TCP BYPASS bound to physical interface {}", idx);
+                            }
+                        }
+                    } else {
+                        tracing::warn!("TUN TCP BYPASS has no physical interface index!");
+                    }
+                    #[cfg(target_os = "linux")]
+                    if let Some(ref name) = lin_name {
+                        let _ = crate::tunnel::proxy::bind_socket_to_interface(&socket, name);
+                    }
+
+                    match tokio::time::timeout(
+                        std::time::Duration::from_secs(10),
+                        socket.connect(remote),
+                    )
+                    .await
+                    {
+                        Ok(Ok(mut direct)) => {
+                            if sniff_len > 0 {
+                                if direct.write_all(&sniff_buf[..sniff_len]).await.is_err() {
+                                    return;
+                                }
+                            }
+                            let _ = tokio::io::copy_bidirectional(&mut stream, &mut direct).await;
+                        }
+                        _ => {
+                            tracing::debug!("Direct bypass connect to {remote} failed");
+                        }
+                    }
+                    return;
+                }
+
+                // ── Tunnel path: forward via local OSTP SOCKS5 proxy ──────────
+                let Ok(mut socks) = tokio::net::TcpStream::connect(&proxy_addr).await else {
+                    return;
+                };
+
+                // SOCKS5 handshake (no auth)
+                if socks.write_all(&[5, 1, 0]).await.is_err() { return; }
+                let mut buf2 = [0u8; 2];
+                if socks.read_exact(&mut buf2).await.is_err() || buf2[0] != 5 || buf2[1] != 0 {
+                    return;
+                }
+
+                // CONNECT request
+                let mut req = vec![5u8, 1, 0];
+                match remote.ip() {
+                    std::net::IpAddr::V4(v4) => {
+                        req.push(1);
+                        req.extend_from_slice(&v4.octets());
+                    }
+                    std::net::IpAddr::V6(v6) => {
+                        req.push(4);
+                        req.extend_from_slice(&v6.octets());
+                    }
+                }
+                req.extend_from_slice(&remote.port().to_be_bytes());
+                if socks.write_all(&req).await.is_err() { return; }
+
+                let mut rep = [0u8; 10];
+                if socks.read_exact(&mut rep).await.is_err() || rep[1] != 0 { return; }
+
+                // Replay sniffed bytes
+                if sniff_len > 0 && socks.write_all(&sniff_buf[..sniff_len]).await.is_err() {
+                    return;
+                }
+
+                let _ = tokio::io::copy_bidirectional(&mut stream, &mut socks).await;
+            });
+        }
+    });
+
+    tracing::info!("NATIVE TUN tunnel active.");
+
+    tokio::select! {
+        _ = shutdown.changed() => {}
+        _ = &mut runner_task => {}
+        _ = &mut tun_to_stack => {}
+        _ = &mut stack_to_tun => {}
+        _ = &mut udp_proxy_task => {}
+        _ = &mut tcp_accept_task => {}
+    }
+
+    tracing::info!("Deactivating NATIVE TUN tunnel...");
+
+    // ── Cleanup ───────────────────────────────────────────────────────────────
+    // Cleanup is handled automatically by the _route_guard Drop trait in ostp-tun
+
+    Ok(())
+}
+
+// ──────────────────────────────────────────────────────────────────────────────
+// Stub for unsupported platforms
+// ──────────────────────────────────────────────────────────────────────────────
+
+#[cfg(not(any(target_os = "windows", target_os = "linux")))]
+pub async fn run_native_tunnel(
+    _config: crate::config::ClientConfig,
+    _shutdown: watch::Receiver<bool>,
+    _exclusions_rx: watch::Receiver<crate::config::ExclusionConfig>,
+) -> Result<()> {
+    Err(anyhow!("Native TUN tunnel is only supported on Windows/Linux"))
+}
+
+// ──────────────────────────────────────────────────────────────────────────────
+// Android: TUN from file-descriptor (opened by VpnService)
+// ──────────────────────────────────────────────────────────────────────────────
+
+#[cfg(target_os = "android")]
+pub async fn run_native_tunnel_from_fd(
+    config: crate::config::ClientConfig,
+    mut shutdown: watch::Receiver<bool>,
+    mut exclusions_rx: watch::Receiver<crate::config::ExclusionConfig>,
+    fd: i32,
+) -> Result<()> {
+    use netstack_smoltcp::StackBuilder;
+    use tokio::io::{AsyncReadExt, AsyncWriteExt};
+    use futures::{StreamExt, SinkExt};
+    use std::os::unix::io::{FromRawFd, AsRawFd};
+
+    let debug = config.debug;
+    tracing::info!("Initializing NATIVE TUN tunnel on Android (FD {})", fd);
+
+    unsafe {
+        let flags = libc::fcntl(fd, libc::F_GETFL);
+        if flags >= 0 {
+            libc::fcntl(fd, libc::F_SETFL, flags | libc::O_NONBLOCK);
+        }
+    }
+
+    let read_fd = unsafe { libc::dup(fd) };
+    if read_fd < 0 {
+        return Err(anyhow!("Failed to dup tun fd for reading"));
+    }
+
+    let file = unsafe { std::fs::File::from_raw_fd(read_fd) };
+    let tun_stream = tokio::io::unix::AsyncFd::new(file)?;
+
+    let (stack, tcp_runner, udp_socket, tcp_listener) = StackBuilder::default()
+        .stack_buffer_size(1024)
+        .tcp_buffer_size(1024)
+        .udp_buffer_size(1024)
+        .enable_tcp(true)
+        .enable_udp(true)
+        .mtu(config.ostp.mtu)
+        .build()?;
+
+    let mut runner_task = tokio::spawn(async move {
+        if let Some(runner) = tcp_runner {
+            let _ = runner.await;
+        }
+    });
+
+    let (mut stack_sink, mut stack_stream) = stack.split();
+
+    let _tun_to_stack = tokio::spawn(async move {
+        let mut buf = vec![0u8; 65536];
+        loop {
+            let mut guard = match tun_stream.readable().await {
+                Ok(g) => g,
+                Err(_) => break,
+            };
+            let n = match guard.try_io(|inner| {
+                let res = unsafe {
+                    libc::read(
+                        inner.as_raw_fd(),
+                        buf.as_mut_ptr() as *mut libc::c_void,
+                        buf.len(),
+                    )
+                };
+                if res < 0 {
+                    let err = std::io::Error::last_os_error();
+                    if err.kind() == std::io::ErrorKind::WouldBlock {
+                        Err(err)
+                    } else {
+                        Ok(0_isize)
+                    }
+                } else {
+                    Ok(res)
+                }
+            }) {
+                Ok(Ok(n)) if n > 0 => n as usize,
+                Ok(Ok(_)) => continue,
+                Ok(Err(_)) => continue,
+                Err(_) => continue,
+            };
+
+            let frame = buf[..n].to_vec();
+            if let Err(e) = stack_sink.send(frame).await {
+                if e.kind() == std::io::ErrorKind::BrokenPipe {
+                    break;
+                }
+            }
+        }
+    });
+
+    let write_fd = unsafe { libc::dup(fd) };
+    if write_fd < 0 {
+        return Err(anyhow!("Failed to dup tun fd for writing"));
+    }
+    unsafe {
+        let flags = libc::fcntl(write_fd, libc::F_GETFL);
+        if flags >= 0 {
+            libc::fcntl(write_fd, libc::F_SETFL, flags | libc::O_NONBLOCK);
+        }
+    }
+    let write_file = unsafe { std::fs::File::from_raw_fd(write_fd) };
+    let tun_write_stream = tokio::io::unix::AsyncFd::new(write_file)?;
+
+    let _stack_to_tun = tokio::spawn(async move {
+        while let Some(Ok(frame)) = stack_stream.next().await {
+            let mut written = 0;
+            while written < frame.len() {
+                let mut guard = match tun_write_stream.writable().await {
+                    Ok(g) => g,
+                    Err(_) => break,
+                };
+                let res = guard.try_io(|inner| {
+                    let res = unsafe {
+                        libc::write(
+                            inner.as_raw_fd(),
+                            frame[written..].as_ptr() as *const libc::c_void,
+                            frame.len() - written,
+                        )
+                    };
+                    if res < 0 {
+                        let err = std::io::Error::last_os_error();
+                        if err.kind() == std::io::ErrorKind::WouldBlock {
+                            Err(err)
+                        } else {
+                            Ok(res)
+                        }
+                    } else {
+                        Ok(res)
+                    }
+                });
+                match res {
+                    Ok(Ok(n)) if n > 0 => written += n as usize,
+                    Ok(Ok(_)) => break,
+                    Ok(Err(_)) => break,
+                    Err(_) => continue,
+                }
+            }
+        }
+    });
+
+    let mut proxy_addr = config.local_proxy.bind_addr.clone();
+    if proxy_addr.starts_with("0.0.0.0:") {
+        proxy_addr = proxy_addr.replace("0.0.0.0:", "127.0.0.1:");
+    }
+
+    let current_exclusions = exclusions_rx.borrow().clone();
+    let matcher = crate::tunnel::exclusion::ExclusionMatcher::new(&current_exclusions, None, None);
+    let matcher_arc = std::sync::Arc::new(tokio::sync::RwLock::new(matcher));
+    
+    let matcher_clone = matcher_arc.clone();
+    tokio::spawn(async move {
+        while let Ok(_) = exclusions_rx.changed().await {
+            let current = exclusions_rx.borrow().clone();
+            let new_matcher = crate::tunnel::exclusion::ExclusionMatcher::new(&current, None, None);
+            *matcher_clone.write().await = new_matcher;
+            if true {
+                tracing::debug!("Android TUN exclusions hot-reloaded");
+            }
+        }
+    });
+
+    let udp_proxy_addr = proxy_addr.clone();
+    let debug_udp = debug;
+    let udp_matcher = matcher_arc.clone();
+    let mut udp_proxy_task = tokio::spawn(async move {
+        if let Some(udp_sock) = udp_socket {
+            super::udp_nat::run_udp_nat(udp_sock, udp_proxy_addr, debug_udp, udp_matcher, None, None).await;
+        }
+    });
+
+
+
+    let mut tcp_accept_task = tokio::spawn(async move {
+        let Some(mut listener) = tcp_listener else { return; };
+
+        while let Some((mut stream, local, remote)) = listener.next().await {
+            let proxy_addr = proxy_addr.clone();
+            let matcher_arc = matcher_arc.clone();
+
+            tokio::spawn(async move {
+                let matcher = matcher_arc.read().await.clone();
+
+                if true {
+                    tracing::debug!("Android TUN TCP {local} → {remote}");
+                }
+
+                // Sniff SNI
+                let mut sniff_buf = [0u8; 2048];
+                let sniff_len =
+                    match tokio::time::timeout(
+                        std::time::Duration::from_millis(100),
+                        stream.read(&mut sniff_buf),
+                    )
+                    .await
+                    {
+                        Ok(Ok(n)) => n,
+                        _ => 0,
+                    };
+
+                let mut should_bypass = false;
+
+                // 1. SNI domain
+                if sniff_len > 0 {
+                    if let Some(sni) =
+                        crate::tunnel::sni_sniff::extract_sni(&sniff_buf[..sniff_len])
+                    {
+                        if true { tracing::debug!("Android TUN SNI: {sni}"); }
+                        if matcher.match_domain(&sni) {
+                            should_bypass = true;
+                        }
+                    }
+                }
+
+                // 2. Process (Android: /proc/net lookup)
+                if !should_bypass {
+                    if let Some(exe) =
+                        crate::tunnel::process_lookup::get_process_name_from_port(local.port())
+                    {
+                        if true {
+                            tracing::debug!("Android TUN port {} → EXE: {}", local.port(), exe);
+                        }
+                        if matcher.match_process(&exe) {
+                            should_bypass = true;
+                        }
+                    }
+                }
+
+                // 3. IP CIDR
+                if !should_bypass && matcher.match_ip(&remote.ip()) {
+                    should_bypass = true;
+                }
+
+                // Bypass: connect directly (Android VPN service already protects the socket
+                // from re-entering the TUN through VpnService.protect())
+                if should_bypass {
+                    if true {
+                        tracing::debug!("Android TUN BYPASS: {remote}");
+                    }
+                    let socket = match remote {
+                        std::net::SocketAddr::V4(_) => tokio::net::TcpSocket::new_v4(),
+                        std::net::SocketAddr::V6(_) => tokio::net::TcpSocket::new_v6(),
+                    };
+                    let Ok(socket) = socket else { return; };
+
+                    match tokio::time::timeout(
+                        std::time::Duration::from_secs(10),
+                        socket.connect(remote),
+                    )
+                    .await
+                    {
+                        Ok(Ok(mut direct)) => {
+                            if sniff_len > 0 {
+                                if direct.write_all(&sniff_buf[..sniff_len]).await.is_err() {
+                                    return;
+                                }
+                            }
+                            let _ = tokio::io::copy_bidirectional(&mut stream, &mut direct).await;
+                        }
+                        _ => {
+                            tracing::debug!("Android bypass connect to {remote} failed");
+                        }
+                    }
+                    return;
+                }
+
+                // Tunnel via SOCKS5 proxy
+                let Ok(mut socks) = tokio::net::TcpStream::connect(&proxy_addr).await else {
+                    return;
+                };
+                if socks.write_all(&[5, 1, 0]).await.is_err() { return; }
+                let mut buf2 = [0u8; 2];
+                if socks.read_exact(&mut buf2).await.is_err() || buf2[0] != 5 || buf2[1] != 0 {
+                    return;
+                }
+                let mut req = vec![5u8, 1, 0];
+                match remote.ip() {
+                    std::net::IpAddr::V4(v4) => {
+                        req.push(1);
+                        req.extend_from_slice(&v4.octets());
+                    }
+                    std::net::IpAddr::V6(v6) => {
+                        req.push(4);
+                        req.extend_from_slice(&v6.octets());
+                    }
+                }
+                req.extend_from_slice(&remote.port().to_be_bytes());
+                if socks.write_all(&req).await.is_err() { return; }
+                let mut rep = [0u8; 10];
+                if socks.read_exact(&mut rep).await.is_err() || rep[1] != 0 { return; }
+                if sniff_len > 0 && socks.write_all(&sniff_buf[..sniff_len]).await.is_err() {
+                    return;
+                }
+                let _ = tokio::io::copy_bidirectional(&mut stream, &mut socks).await;
+            });
+        }
+    });
+
+    tracing::info!("NATIVE TUN (Android) tunnel active.");
+
+    tokio::select! {
+        _ = shutdown.changed() => {}
+        _ = &mut runner_task => {}
+        _ = _tun_to_stack => {}
+        _ = _stack_to_tun => {}
+        _ = &mut udp_proxy_task => {}
+        _ = &mut tcp_accept_task => {}
+    }
+
+    tracing::info!("NATIVE TUN (Android) deactivated.");
+    Ok(())
+}
+
+#[cfg(not(target_os = "android"))]
+pub async fn run_native_tunnel_from_fd(
+    _config: crate::config::ClientConfig,
+    _shutdown: watch::Receiver<bool>,
+    _exclusions_rx: watch::Receiver<crate::config::ExclusionConfig>,
+    _fd: i32,
+) -> Result<()> {
+    Err(anyhow!("Native TUN from FD is only supported on Android"))
+}

ostp-client/src/tunnel/process_lookup.rs

@@ -0,0 +1,194 @@
+#[cfg(target_os = "windows")]
+pub fn get_process_name_from_port(port: u16) -> Option<String> {
+    use winapi::shared::minwindef::ULONG;
+    use winapi::shared::winerror::ERROR_INSUFFICIENT_BUFFER;
+    use winapi::um::iphlpapi::GetExtendedTcpTable;
+    use winapi::shared::tcpmib::{MIB_TCPTABLE_OWNER_PID, MIB_TCPROW_OWNER_PID};
+
+    let mut size: ULONG = 0;
+    let table_class = 5; // TCP_TABLE_OWNER_PID_ALL
+    let mut table = vec![0u8; 1024];
+
+    unsafe {
+        let mut ret = GetExtendedTcpTable(
+            table.as_mut_ptr() as *mut _,
+            &mut size,
+            0,
+            2, // AF_INET
+            table_class,
+            0,
+        );
+
+        if ret == ERROR_INSUFFICIENT_BUFFER {
+            table.resize(size as usize, 0);
+            ret = GetExtendedTcpTable(
+                table.as_mut_ptr() as *mut _,
+                &mut size,
+                0,
+                2, // AF_INET
+                table_class,
+                0,
+            );
+        }
+
+        if ret == 0 {
+            let tcp_table = &*(table.as_ptr() as *const MIB_TCPTABLE_OWNER_PID);
+            let row_ptr = &tcp_table.table[0] as *const MIB_TCPROW_OWNER_PID;
+            for i in 0..tcp_table.dwNumEntries {
+                let row = &*row_ptr.add(i as usize);
+                // Local port is in network byte order
+                let local_port = u16::from_be(row.dwLocalPort as u16);
+                if local_port == port {
+                    return get_process_name_from_pid(row.dwOwningPid);
+                }
+            }
+        }
+    }
+    None
+}
+
+#[cfg(target_os = "windows")]
+pub fn get_process_name_from_port_udp(port: u16) -> Option<String> {
+    use winapi::shared::minwindef::ULONG;
+    use winapi::shared::winerror::ERROR_INSUFFICIENT_BUFFER;
+    use winapi::um::iphlpapi::GetExtendedUdpTable;
+    use winapi::shared::udpmib::{MIB_UDPTABLE_OWNER_PID, MIB_UDPROW_OWNER_PID};
+
+    let mut size: ULONG = 0;
+    let table_class = 1; // UDP_TABLE_OWNER_PID
+    let mut table = vec![0u8; 1024];
+
+    unsafe {
+        let mut ret = GetExtendedUdpTable(
+            table.as_mut_ptr() as *mut _,
+            &mut size,
+            0,
+            2, // AF_INET
+            table_class,
+            0,
+        );
+
+        if ret == ERROR_INSUFFICIENT_BUFFER {
+            table.resize(size as usize, 0);
+            ret = GetExtendedUdpTable(
+                table.as_mut_ptr() as *mut _,
+                &mut size,
+                0,
+                2, // AF_INET
+                table_class,
+                0,
+            );
+        }
+
+        if ret == 0 {
+            let udp_table = &*(table.as_ptr() as *const MIB_UDPTABLE_OWNER_PID);
+            let row_ptr = &udp_table.table[0] as *const MIB_UDPROW_OWNER_PID;
+            for i in 0..udp_table.dwNumEntries {
+                let row = &*row_ptr.add(i as usize);
+                let local_port = u16::from_be(row.dwLocalPort as u16);
+                if local_port == port {
+                    return get_process_name_from_pid(row.dwOwningPid);
+                }
+            }
+        }
+    }
+    None
+}
+
+#[cfg(target_os = "windows")]
+fn get_process_name_from_pid(pid: u32) -> Option<String> {
+    use winapi::um::processthreadsapi::OpenProcess;
+    use winapi::um::psapi::GetModuleBaseNameW;
+    use winapi::um::winnt::{PROCESS_QUERY_INFORMATION, PROCESS_VM_READ};
+    use winapi::um::handleapi::CloseHandle;
+    use std::os::windows::ffi::OsStringExt;
+
+    unsafe {
+        let handle = OpenProcess(PROCESS_QUERY_INFORMATION | PROCESS_VM_READ, 0, pid);
+        if handle.is_null() {
+            return None;
+        }
+
+        let mut buffer = [0u16; 1024];
+        let len = GetModuleBaseNameW(handle, std::ptr::null_mut(), buffer.as_mut_ptr(), buffer.len() as u32);
+        CloseHandle(handle);
+
+        if len > 0 {
+            let name = std::ffi::OsString::from_wide(&buffer[..len as usize]);
+            return Some(name.to_string_lossy().into_owned());
+        }
+    }
+    None
+}
+
+#[cfg(target_os = "linux")]
+pub fn get_process_name_from_port(port: u16) -> Option<String> {
+    use std::fs;
+    use std::io::{BufRead, BufReader};
+
+    let hex_port = format!("{:04X}", port);
+
+    let check_net_file = |path: &str| -> Option<u64> {
+        let file = fs::File::open(path).ok()?;
+        let reader = BufReader::new(file);
+        for line in reader.lines().skip(1).filter_map(Result::ok) {
+            let parts: Vec<&str> = line.split_whitespace().collect();
+            if parts.len() >= 10 {
+                let local_addr = parts[1];
+                if local_addr.ends_with(&format!(":{}", hex_port)) {
+                    if let Ok(inode) = parts[9].parse::<u64>() {
+                        return Some(inode);
+                    }
+                }
+            }
+        }
+        None
+    };
+
+    let target_inode = check_net_file("/proc/net/tcp")
+        .or_else(|| check_net_file("/proc/net/tcp6"))
+        .or_else(|| check_net_file("/proc/net/udp"))
+        .or_else(|| check_net_file("/proc/net/udp6"));
+
+    let target_inode = target_inode?;
+    let socket_str = format!("socket:[{}]", target_inode);
+
+    for entry in fs::read_dir("/proc").ok()?.filter_map(Result::ok) {
+        let file_name = entry.file_name();
+        let pid_str = file_name.to_string_lossy();
+        if !pid_str.chars().all(char::is_numeric) {
+            continue;
+        }
+
+        let fd_dir = entry.path().join("fd");
+        if let Ok(fd_entries) = fs::read_dir(fd_dir) {
+            for fd_entry in fd_entries.filter_map(Result::ok) {
+                if let Ok(target) = fs::read_link(fd_entry.path()) {
+                    if target.to_string_lossy() == socket_str {
+                        let exe_path = entry.path().join("exe");
+                        if let Ok(exe_link) = fs::read_link(exe_path) {
+                            if let Some(name) = exe_link.file_name() {
+                                return Some(name.to_string_lossy().into_owned());
+                            }
+                        }
+                        if let Ok(comm) = fs::read_to_string(entry.path().join("comm")) {
+                            return Some(comm.trim().to_string());
+                        }
+                    }
+                }
+            }
+        }
+    }
+
+    None
+}
+
+#[cfg(not(any(target_os = "windows", target_os = "linux")))]
+pub fn get_process_name_from_port(_port: u16) -> Option<String> {
+    None
+}
+
+#[cfg(not(target_os = "windows"))]
+pub fn get_process_name_from_port_udp(port: u16) -> Option<String> {
+    get_process_name_from_port(port)
+}

ostp-client/src/tunnel/proxy.rs

@@ -1,37 +1,232 @@
 use std::collections::HashMap;
+use crate::tunnel::exclusion::ExclusionMatcher;
 use anyhow::{anyhow, Context, Result};
 use tokio::io::{AsyncReadExt, AsyncWriteExt};
-use tokio::net::{TcpListener, TcpStream};
+use tokio::net::{TcpListener, TcpStream, UdpSocket};
+use std::sync::Arc;
 use tokio::sync::{mpsc, watch};
 use tokio::time::{timeout, Duration};
 
 use crate::config::{ExclusionConfig, LocalProxyConfig, OstpConfig};
 use crate::tunnel::{ProxyEvent, ProxyToClientMsg};
 
+#[cfg(target_os = "windows")]
+use std::os::windows::io::AsRawSocket;
+
+#[cfg(target_os = "linux")]
+use std::os::fd::AsRawFd;
+
+#[cfg(target_os = "windows")]
+#[link(name = "ws2_32")]
+extern "system" {
+    fn setsockopt(
+        s: usize,
+        level: i32,
+        optname: i32,
+        optval: *const u8,
+        optlen: i32,
+    ) -> i32;
+}
+
+#[cfg(target_os = "windows")]
+pub fn bind_socket_to_interface(socket: &impl AsRawSocket, is_ipv6: bool, if_index: u32) -> std::io::Result<()> {
+    let s = socket.as_raw_socket() as usize;
+    if is_ipv6 {
+        // IPV6_UNICAST_IF expects interface index in host byte order
+        let optval = if_index;
+        let ret = unsafe {
+            setsockopt(
+                s,
+                41, // IPPROTO_IPV6
+                31, // IPV6_UNICAST_IF
+                &optval as *const u32 as *const u8,
+                4,
+            )
+        };
+        if ret != 0 {
+            return Err(std::io::Error::last_os_error());
+        }
+    } else {
+        // IP_UNICAST_IF expects interface index in NETWORK byte order (big-endian)
+        let optval = if_index.to_be();
+        let ret = unsafe {
+            setsockopt(
+                s,
+                0, // IPPROTO_IP
+                31, // IP_UNICAST_IF
+                &optval as *const u32 as *const u8,
+                4,
+            )
+        };
+        if ret != 0 {
+            return Err(std::io::Error::last_os_error());
+        }
+    }
+    Ok(())
+}
+
+#[cfg(target_os = "linux")]
+pub fn bind_socket_to_interface(socket: &impl AsRawFd, if_name: &str) -> std::io::Result<()> {
+    let fd = socket.as_raw_fd();
+    let mut if_name_bytes = if_name.as_bytes().to_vec();
+    if_name_bytes.push(0);
+    let ret = unsafe {
+        libc::setsockopt(
+            fd,
+            libc::SOL_SOCKET,
+            libc::SO_BINDTODEVICE,
+            if_name_bytes.as_ptr() as *const std::ffi::c_void,
+            if_name_bytes.len() as libc::socklen_t,
+        )
+    };
+    if ret != 0 {
+        return Err(std::io::Error::last_os_error());
+    }
+    Ok(())
+}
+
+pub fn get_windows_physical_if_index() -> Option<u32> {
+    #[cfg(target_os = "windows")]
+    {
+        return ostp_tun::windows::windows_route::sys::get_default_ipv4_route().map(|(_, idx)| idx);
+    }
+    #[cfg(not(target_os = "windows"))]
+    {
+        None
+    }
+}
+
+pub fn get_linux_physical_if_name() -> Option<String> {
+    #[cfg(target_os = "linux")]
+    {
+        let output = std::process::Command::new("ip")
+            .args(["route", "show", "default"])
+            .output()
+            .ok()?;
+        if output.status.success() {
+            let s = String::from_utf8_lossy(&output.stdout);
+            if let Some(dev_part) = s.split_whitespace().skip_while(|w| *w != "dev").nth(1) {
+                return Some(dev_part.to_string());
+            }
+        }
+    }
+    None
+}
+
+#[allow(unused_variables)]
+async fn connect_bypassing_tun(
+    target: &str,
+    physical_if_index: Option<u32>,
+    _physical_if_name: &Option<String>,
+) -> Result<TcpStream> {
+    let resolved = tokio::net::lookup_host(target).await
+        .with_context(|| format!("failed to resolve host for bypass connect: {target}"))?;
+
+    let mut last_err = None;
+    for addr in resolved {
+        let socket = if addr.is_ipv6() {
+            let s = tokio::net::TcpSocket::new_v6()?;
+            let _ = s.bind("[::]:0".parse().unwrap());
+            s
+        } else {
+            let s = tokio::net::TcpSocket::new_v4()?;
+            let _ = s.bind("0.0.0.0:0".parse().unwrap());
+            s
+        };
+
+        #[cfg(target_os = "windows")]
+        if let Some(if_index) = physical_if_index {
+            if let Err(e) = bind_socket_to_interface(&socket, addr.is_ipv6(), if_index) {
+                tracing::warn!("Failed to bind TCP socket to interface {}: {}", if_index, e);
+            }
+        }
+
+        #[cfg(target_os = "linux")]
+        if let Some(ref if_name) = _physical_if_name {
+            if let Err(e) = bind_socket_to_interface(&socket, if_name) {
+                tracing::warn!("Failed to bind TCP socket to interface {}: {}", if_name, e);
+            }
+        }
+
+        match socket.connect(addr).await {
+            Ok(stream) => return Ok(stream),
+            Err(e) => {
+                last_err = Some(e);
+            }
+        }
+    }
+
+    Err(anyhow!(
+        "direct connect failed: {:?}",
+        last_err.map(|e| e.to_string()).unwrap_or_else(|| "no addresses resolved".to_string())
+    ))
+}
+
+#[allow(unused_variables)]
+async fn create_udp_socket_bypassing_tun(
+    is_ipv6: bool,
+    physical_if_index: Option<u32>,
+    _physical_if_name: &Option<String>,
+) -> Result<UdpSocket> {
+    let addr: std::net::SocketAddr = if is_ipv6 {
+        "[::]:0".parse().unwrap()
+    } else {
+        "0.0.0.0:0".parse().unwrap()
+    };
+    
+    let socket = UdpSocket::bind(addr).await
+        .with_context(|| format!("failed to bind direct UdpSocket to wildcard {}", addr))?;
+
+    #[cfg(target_os = "windows")]
+    if let Some(if_index) = physical_if_index {
+        if let Err(e) = bind_socket_to_interface(&socket, is_ipv6, if_index) {
+            tracing::warn!("Failed to bind UDP socket to interface index {}: {}", if_index, e);
+        }
+    }
+
+    #[cfg(target_os = "linux")]
+    if let Some(ref if_name) = _physical_if_name {
+        if let Err(e) = bind_socket_to_interface(&socket, if_name) {
+            tracing::warn!("Failed to bind UDP socket to interface {}: {}", if_name, e);
+        }
+    }
+
+    Ok(socket)
+}
+
 pub async fn run_local_socks5_proxy(
     cfg: LocalProxyConfig,
-    _ostp: OstpConfig,
-    exclusions: ExclusionConfig,
+    ostp: OstpConfig,
+    mut exclusions_rx: watch::Receiver<ExclusionConfig>,
     debug: bool,
     mut shutdown: watch::Receiver<bool>,
     proxy_events_tx: mpsc::Sender<ProxyEvent>,
-    mut client_msgs_rx: mpsc::Receiver<(u16, ProxyToClientMsg)>,
+    mut client_msgs_rx: mpsc::UnboundedReceiver<(u16, ProxyToClientMsg)>,
 ) -> Result<()> {
     let connect_timeout = Duration::from_millis(cfg.connect_timeout_ms.max(1));
     let listener = TcpListener::bind(&cfg.bind_addr)
         .await
         .with_context(|| format!("failed to bind local HTTP/SOCKS5 proxy at {}", cfg.bind_addr))?;
 
-    if debug {
-        eprintln!("[ostp-client] local HTTP/SOCKS5 proxy listening at {}", cfg.bind_addr);
-        eprintln!("[ostp-client] Windows system proxy: set HTTP proxy to {}. tun2socks: SOCKS5 on same address.", cfg.bind_addr);
+    tracing::info!("local HTTP/SOCKS5 proxy listening at {}", cfg.bind_addr);
+
+    let physical_if_index = tokio::task::spawn_blocking(get_windows_physical_if_index).await.unwrap_or(None);
+    let physical_if_name = tokio::task::spawn_blocking(get_linux_physical_if_name).await.unwrap_or(None);
+
+    if physical_if_index.is_some() {
+        tracing::info!("Local proxy physical interface index: {:?}", physical_if_index);
+    }
+    if physical_if_name.is_some() {
+        tracing::info!("Local proxy physical interface name: {:?}", physical_if_name);
     }
 
-    let matcher = ExclusionMatcher::new(&exclusions);
+    let mut current_exclusions = exclusions_rx.borrow().clone();
+    let mut matcher = ExclusionMatcher::new(&current_exclusions, physical_if_index, physical_if_name.clone());
     let (connect_tx, mut connect_rx) = mpsc::channel(128);
+    let max_chunk = ostp.mtu.saturating_sub(150).max(512);
 
     let mut next_stream_id: u16 = 1;
-    let mut active_streams: HashMap<u16, mpsc::Sender<ProxyToClientMsg>> = HashMap::new();
+    let mut active_streams: HashMap<u16, mpsc::UnboundedSender<ProxyToClientMsg>> = HashMap::new();
 
     loop {
         tokio::select! {
@@ -40,13 +235,24 @@ pub async fn run_local_socks5_proxy(
                     break;
                 }
             }
+            Ok(_) = exclusions_rx.changed() => {
+                current_exclusions = exclusions_rx.borrow().clone();
+                matcher = ExclusionMatcher::new(&current_exclusions, physical_if_index, physical_if_name.clone());
+                if true {
+                    tracing::info!("Local proxy exclusions hot-reloaded");
+                }
+            }
             accepted = listener.accept() => {
                 let (socket, _) = accepted?;
                 let stream_id = next_stream_id;
-                next_stream_id = next_stream_id.wrapping_add(1);
-                if next_stream_id == 0 { next_stream_id = 1; }
+                // Advance, skipping zero and any stream_id still in active_streams
+                loop {
+                    next_stream_id = next_stream_id.wrapping_add(1);
+                    if next_stream_id == 0 { next_stream_id = 1; }
+                    if !active_streams.contains_key(&next_stream_id) { break; }
+                }
 
-                let (tx, rx) = mpsc::channel(256);
+                let (tx, rx) = mpsc::unbounded_channel();
                 active_streams.insert(stream_id, tx);
 
                 let event_tx = proxy_events_tx.clone();
@@ -62,6 +268,7 @@ pub async fn run_local_socks5_proxy(
                         connect_timeout,
                         debug,
                         matcher_clone,
+                        max_chunk,
                     ).await {
                         let msg = err.to_string();
                         // Suppress routine disconnects and unsupported SOCKS5 command attempts (like UDP) from spam logs
@@ -69,17 +276,24 @@ pub async fn run_local_socks5_proxy(
                             && !msg.contains("Connection reset")
                             && !msg.contains("Broken pipe")
                             && !msg.contains("unsupported SOCKS5 command")
-                        {
-                            if debug {
-                                eprintln!("[ostp-client] proxy client error: {err}");
+                            && debug {
+                                tracing::warn!("proxy client error: {err}");
                             }
-                        }
                     }
                 });
             }
             Some((stream_id, msg)) = client_msgs_rx.recv() => {
-                if let Some(tx) = active_streams.get(&stream_id) {
-                    if tx.send(msg).await.is_err() {
+                if stream_id == 0 {
+                    if let ProxyToClientMsg::Close = msg {
+                        if true {
+                            tracing::info!("Resetting all active proxy streams on reconnect");
+                        }
+                        for (_, tx) in active_streams.drain() {
+                            let _ = tx.send(ProxyToClientMsg::Close);
+                        }
+                    }
+                } else if let Some(tx) = active_streams.get(&stream_id) {
+                    if tx.send(msg).is_err() {
                         active_streams.remove(&stream_id);
                     }
                 }
@@ -112,16 +326,275 @@ fn extract_host_port(uri: &str, default_port: u16) -> String {
     }
 }
 
+struct StreamGuard {
+    stream_id: u16,
+    close_tx: mpsc::Sender<u16>,
+}
+
+impl Drop for StreamGuard {
+    fn drop(&mut self) {
+        let tx = self.close_tx.clone();
+        let id = self.stream_id;
+        tokio::spawn(async move {
+            let _ = tx.send(id).await;
+        });
+    }
+}
+
+async fn handle_udp_associate(
+    mut client_tcp: TcpStream,
+    udp_socket: tokio::net::UdpSocket,
+    stream_id: u16,
+    event_tx: mpsc::Sender<ProxyEvent>,
+    mut rx: mpsc::UnboundedReceiver<ProxyToClientMsg>,
+    close_tx: mpsc::Sender<u16>,
+    debug: bool,
+    matcher: ExclusionMatcher,
+    connect_timeout: Duration,
+) -> Result<()> {
+    let client_udp_addr = Arc::new(std::sync::Mutex::new(None));
+    let mut buf = vec![0u8; 65536];
+    
+    let udp_socket = Arc::new(udp_socket);
+    let sock_rx = udp_socket.clone();
+    let sock_tx = udp_socket;
+
+    let mut direct_udp_v4: Option<Arc<UdpSocket>> = None;
+    let mut direct_udp_v6: Option<Arc<UdpSocket>> = None;
+    // Held only to keep the direct-UDP readers' cancellation senders alive;
+    // dropping this (on every return path from this function) is what tells
+    // spawn_direct_udp_reader's tasks to stop. See its doc comment.
+    let mut direct_udp_cancel_txs: Vec<tokio::sync::oneshot::Sender<()>> = Vec::new();
+
+    let mut tcp_buf = [0u8; 1];
+    loop {
+        tokio::select! {
+            res = client_tcp.read(&mut tcp_buf) => {
+                match res {
+                    Ok(0) | Err(_) => break,
+                    Ok(_) => {}
+                }
+            }
+            res = sock_rx.recv_from(&mut buf) => {
+                let (len, addr) = match res {
+                    Ok(v) => v,
+                    Err(e) => {
+                        tracing::debug!("udp_associate recv_from error: {}", e);
+                        continue; // transient error, don't kill the session
+                    }
+                };
+                {
+                    let mut guard = client_udp_addr.lock().unwrap();
+                    if guard.is_none() {
+                        *guard = Some(addr);
+                    }
+                }
+                if len < 4 { continue; }
+                let frag = buf[2];
+                if frag != 0 { continue; } // Fragmented UDP not supported
+                let atyp = buf[3];
+                let (header_len, target) = match atyp {
+                    0x01 => {
+                        if len < 10 { continue; }
+                        let ip = std::net::Ipv4Addr::new(buf[4], buf[5], buf[6], buf[7]);
+                        let port = u16::from_be_bytes([buf[8], buf[9]]);
+                        (10, format!("{}:{}", ip, port))
+                    }
+                    0x03 => {
+                        if len < 5 { continue; }
+                        let domain_len = buf[4] as usize;
+                        if len < 5 + domain_len + 2 { continue; }
+                        let domain = String::from_utf8_lossy(&buf[5..5+domain_len]);
+                        let port = u16::from_be_bytes([buf[5+domain_len], buf[5+domain_len+1]]);
+                        (5 + domain_len + 2, format!("{}:{}", domain, port))
+                    }
+                    0x04 => {
+                        if len < 22 { continue; }
+                        let mut octets = [0u8; 16];
+                        octets.copy_from_slice(&buf[4..20]);
+                        let ip = std::net::Ipv6Addr::from(octets);
+                        let port = u16::from_be_bytes([buf[20], buf[21]]);
+                        (22, format!("[{}]:{}", ip, port))
+                    }
+                    _ => continue,
+                };
+                let payload = bytes::Bytes::copy_from_slice(&buf[header_len..len]);
+
+                let target_host = if let Some((host, _)) = split_host_port(&target) { host } else { target.clone() };
+                let target_port = match split_host_port(&target) { Some((_, p)) => p, None => 0 };
+                // Check if target should bypass the tunnel
+                if matcher.should_bypass_target(&target_host, target_port, connect_timeout).await {
+                    if true {
+                        tracing::debug!("proxy UDP BYPASS target={}", target);
+                    }
+                    // Resolve target to find if it is IPv4 or IPv6
+                    if let Ok(resolved_addrs) = tokio::net::lookup_host(&target).await {
+                        if let Some(target_addr) = resolved_addrs.into_iter().next() {
+                            let is_ipv6 = target_addr.is_ipv6();
+                            let direct_socket = if is_ipv6 {
+                                if direct_udp_v6.is_none() {
+                                    match create_udp_socket_bypassing_tun(true, matcher.physical_if_index, &matcher.physical_if_name).await {
+                                        Ok(s) => {
+                                            let s_arc = Arc::new(s);
+                                            let (cancel_tx, cancel_rx) = tokio::sync::oneshot::channel();
+                                            spawn_direct_udp_reader(s_arc.clone(), sock_tx.clone(), client_udp_addr.clone(), debug, cancel_rx);
+                                            direct_udp_cancel_txs.push(cancel_tx);
+                                            direct_udp_v6 = Some(s_arc);
+                                        }
+                                        Err(e) => {
+                                            tracing::error!("Failed to create bypass UDP v6 socket: {}", e);
+                                        }
+                                    }
+                                }
+                                &direct_udp_v6
+                            } else {
+                                if direct_udp_v4.is_none() {
+                                    match create_udp_socket_bypassing_tun(false, matcher.physical_if_index, &matcher.physical_if_name).await {
+                                        Ok(s) => {
+                                            let s_arc = Arc::new(s);
+                                            let (cancel_tx, cancel_rx) = tokio::sync::oneshot::channel();
+                                            spawn_direct_udp_reader(s_arc.clone(), sock_tx.clone(), client_udp_addr.clone(), debug, cancel_rx);
+                                            direct_udp_cancel_txs.push(cancel_tx);
+                                            direct_udp_v4 = Some(s_arc);
+                                        }
+                                        Err(e) => {
+                                            tracing::error!("Failed to create bypass UDP v4 socket: {}", e);
+                                        }
+                                    }
+                                }
+                                &direct_udp_v4
+                            };
+
+                            if let Some(s) = direct_socket {
+                                if let Err(e) = s.send_to(&payload, target_addr).await {
+                                    if true {
+                                        tracing::warn!("failed to send bypass UDP packet to {}: {}", target_addr, e);
+                                    }
+                                }
+                            }
+                        }
+                    }
+                } else {
+                    tracing::debug!("proxy.rs forwarding UDP DATA to server for target={} payload len={}", target, payload.len());
+                    let _ = event_tx.send(ProxyEvent::UdpData { stream_id, target, payload }).await;
+                }
+            }
+            msg = rx.recv() => {
+                match msg {
+                    Some(ProxyToClientMsg::UdpData(target, data)) => {
+                        if let Some(client_addr) = {
+                            let guard = client_udp_addr.lock().unwrap();
+                            *guard
+                        } {
+                            let mut packet = vec![0x00, 0x00, 0x00];
+                            let mut parts = target.rsplitn(2, ':');
+                            let port_str = parts.next().unwrap_or("0");
+                            let host_str = parts.next().unwrap_or(&target);
+                            let host_str = host_str.trim_start_matches('[').trim_end_matches(']');
+                            let port = port_str.parse::<u16>().unwrap_or(0);
+                            
+                            if let Ok(ipv4) = host_str.parse::<std::net::Ipv4Addr>() {
+                                packet.push(0x01);
+                                packet.extend_from_slice(&ipv4.octets());
+                            } else if let Ok(ipv6) = host_str.parse::<std::net::Ipv6Addr>() {
+                                packet.push(0x04);
+                                packet.extend_from_slice(&ipv6.octets());
+                            } else {
+                                packet.push(0x03);
+                                let bytes = host_str.as_bytes();
+                                packet.push(bytes.len() as u8);
+                                packet.extend_from_slice(bytes);
+                            }
+                            packet.extend_from_slice(&port.to_be_bytes());
+                            packet.extend_from_slice(&data);
+                            tracing::debug!("proxy.rs forwarding UDP REPLY to client_addr={} from server for target={} payload len={}", client_addr, target, data.len());
+                            let _ = sock_tx.send_to(&packet, client_addr).await;
+                        } else {
+                            tracing::error!("proxy.rs failed to parse target string as SocketAddr: {}", target);
+                        }
+                    }
+                    Some(ProxyToClientMsg::Close) | Some(ProxyToClientMsg::Error(_)) | None => break,
+                    _ => {}
+                }
+            }
+        }
+    }
+    let _ = close_tx.send(stream_id).await;
+    Ok(())
+}
+
+fn spawn_direct_udp_reader(
+    direct_socket: Arc<UdpSocket>,
+    sock_tx: Arc<UdpSocket>,
+    client_udp_addr: Arc<std::sync::Mutex<Option<std::net::SocketAddr>>>,
+    _debug: bool,
+    mut cancel_rx: tokio::sync::oneshot::Receiver<()>,
+) {
+    tokio::spawn(async move {
+        let mut buf = vec![0u8; 65536];
+        loop {
+            let recv_result = tokio::select! {
+                // Fires as soon as the sender half (held by handle_udp_associate
+                // for exactly this reason) is dropped - which happens the
+                // instant that function returns, on every exit path, with no
+                // explicit signaling needed. Without this, a UDP-associate
+                // session that ever bypassed traffic direct (excluded IP/
+                // domain) leaked this socket + task for the rest of the
+                // process's life once the session ended: nothing else ever
+                // stopped this loop.
+                _ = &mut cancel_rx => break,
+                res = direct_socket.recv_from(&mut buf) => res,
+            };
+            match recv_result {
+                Ok((len, target_addr)) => {
+                    let client_addr = {
+                        let guard = client_udp_addr.lock().unwrap();
+                        *guard
+                    };
+                    if let Some(client_addr) = client_addr {
+                        let mut packet = vec![0x00, 0x00, 0x00];
+                        if let Ok(ipv4) = target_addr.ip().to_string().parse::<std::net::Ipv4Addr>() {
+                            packet.push(0x01);
+                            packet.extend_from_slice(&ipv4.octets());
+                        } else if let Ok(ipv6) = target_addr.ip().to_string().parse::<std::net::Ipv6Addr>() {
+                            packet.push(0x04);
+                            packet.extend_from_slice(&ipv6.octets());
+                        } else {
+                            continue;
+                        }
+                        packet.extend_from_slice(&target_addr.port().to_be_bytes());
+                        packet.extend_from_slice(&buf[..len]);
+                        if let Err(e) = sock_tx.send_to(&packet, client_addr).await {
+                            if true {
+                                tracing::warn!("failed to send direct UDP response to client: {e}");
+                            }
+                        }
+                    }
+                }
+                Err(e) => {
+                    if true {
+                        tracing::debug!("direct UDP socket read loop exiting: {e}");
+                    }
+                    break;
+                }
+            }
+        }
+    });
+}
+
 async fn handle_proxy_client(
     mut client: TcpStream,
     stream_id: u16,
     event_tx: mpsc::Sender<ProxyEvent>,
-    mut rx: mpsc::Receiver<ProxyToClientMsg>,
+    mut rx: mpsc::UnboundedReceiver<ProxyToClientMsg>,
     close_tx: mpsc::Sender<u16>,
     connect_timeout: Duration,
     debug: bool,
     matcher: ExclusionMatcher,
+    max_chunk: usize,
 ) -> Result<()> {
+    let _guard = StreamGuard { stream_id, close_tx: close_tx.clone() };
+
     // Peek the first byte to distinguish SOCKS5 (0x05) from HTTP (any printable ASCII)
     let mut first_byte = [0_u8; 1];
     client.read_exact(&mut first_byte).await?;
@@ -147,8 +620,10 @@ async fn handle_proxy_client(
         if req[0] != 0x05 {
             return Err(anyhow!("SOCKS5 request version mismatch"));
         }
-        if req[1] != 0x01 {
-            // Not CONNECT — send COMMAND NOT SUPPORTED
+        
+        let is_udp = req[1] == 0x03;
+        if req[1] != 0x01 && !is_udp {
+            // Not CONNECT and Not UDP ASSOCIATE — send COMMAND NOT SUPPORTED
             client.write_all(&[0x05, 0x07, 0x00, 0x01, 0, 0, 0, 0, 0, 0]).await?;
             return Err(anyhow!("unsupported SOCKS5 command {}", req[1]));
         }
@@ -186,11 +661,41 @@ async fn handle_proxy_client(
             }
         };
 
-        if debug {
-            eprintln!("[ostp-client] proxy CONNECT stream_id={stream_id} target={target}");
+        if is_udp {
+            if true { tracing::debug!("proxy UDP ASSOCIATE stream_id={stream_id}"); }
+            let udp_socket = UdpSocket::bind("127.0.0.1:0").await?;
+            let port = udp_socket.local_addr()?.port();
+            let mut reply = vec![0x05, 0x00, 0x00, 0x01, 127, 0, 0, 1];
+            reply.extend_from_slice(&port.to_be_bytes());
+            client.write_all(&reply).await?;
+            
+            event_tx.send(ProxyEvent::UdpAssociate { stream_id }).await?;
+            return handle_udp_associate(
+                client,
+                udp_socket,
+                stream_id,
+                event_tx,
+                rx,
+                close_tx,
+                debug,
+                matcher,
+                connect_timeout,
+            ).await;
         }
-        if matcher.should_bypass(&target, connect_timeout).await {
-            return direct_connect_socks5(client, stream_id, &target, close_tx, debug).await;
+
+        tracing::debug!("proxy CONNECT stream_id={stream_id} target={target}");
+        let target_host = if let Some((host, _)) = split_host_port(&target) { host } else { target.clone() };
+        let target_port = match split_host_port(&target) { Some((_, p)) => p, None => 0 };
+        if matcher.should_bypass_target(&target_host, target_port, connect_timeout).await {
+            return direct_connect_socks5(
+                client,
+                stream_id,
+                &target,
+                matcher.physical_if_index,
+                &matcher.physical_if_name,
+                close_tx,
+                debug,
+            ).await;
         }
         event_tx.send(ProxyEvent::NewStream { stream_id, target: target.clone() }).await?;
 
@@ -220,12 +725,18 @@ async fn handle_proxy_client(
         // Read the rest of the HTTP request headers byte-by-byte
         let mut header_bytes = Vec::with_capacity(512);
         header_bytes.push(first_byte[0]);
-        let mut byte = [0_u8; 1];
+        let mut chunk = [0_u8; 512];
         loop {
-            client.read_exact(&mut byte).await?;
-            header_bytes.push(byte[0]);
-            if header_bytes.ends_with(b"\r\n\r\n") {
-                break;
+            let n = client.read(&mut chunk).await?;
+            if n == 0 {
+                return Err(anyhow!("connection closed during HTTP header read"));
+            }
+            header_bytes.extend_from_slice(&chunk[..n]);
+            if header_bytes.len() >= 4 {
+                let tail = &header_bytes[header_bytes.len().saturating_sub(4)..];
+                if tail.ends_with(b"\r\n\r\n") {
+                    break;
+                }
             }
             if header_bytes.len() > 8192 {
                 client.write_all(b"HTTP/1.1 431 Request Header Fields Too Large\r\n\r\n").await?;
@@ -257,16 +768,20 @@ async fn handle_proxy_client(
             extract_host_port(raw_uri, default_port)
         };
 
-        if debug {
-            eprintln!("[ostp-client] proxy CONNECT stream_id={stream_id} target={target}");
+        if true {
+            tracing::info!("proxy CONNECT stream_id={stream_id} target={target}");
         }
-        if matcher.should_bypass(&target, connect_timeout).await {
+        let target_host = if let Some((host, _)) = split_host_port(&target) { host } else { target.clone() };
+        let target_port = match split_host_port(&target) { Some((_, p)) => p, None => 443 };
+        if matcher.should_bypass_target(&target_host, target_port, connect_timeout).await {
             return direct_connect_http(
                 client,
                 stream_id,
                 &target,
                 method.as_str(),
                 header_bytes,
+                matcher.physical_if_index,
+                &matcher.physical_if_name,
                 close_tx,
                 debug,
             ).await;
@@ -306,28 +821,33 @@ async fn handle_proxy_client(
     }
 
     // ── Bidirectional raw data forwarding ─────────────────────────────
-    let mut tcp_buf = vec![0_u8; 1024];
+    let mut tcp_buf = vec![0_u8; 65536];
     loop {
         tokio::select! {
             read_res = client.read(&mut tcp_buf) => {
                 match read_res {
                     Ok(0) => {
                         let _ = event_tx.send(ProxyEvent::Close { stream_id }).await;
-                        if debug {
-                            eprintln!("[ostp-client] proxy CLOSE stream_id={stream_id}");
+                        if true {
+                            tracing::info!("proxy CLOSE stream_id={stream_id}");
                         }
                         break;
                     }
                     Ok(n) => {
-                        let _ = event_tx.send(ProxyEvent::Data {
-                            stream_id,
-                            payload: bytes::Bytes::copy_from_slice(&tcp_buf[..n]),
-                        }).await;
+                        let mut offset = 0;
+                        while offset < n {
+                            let end = (offset + max_chunk).min(n);
+                            let _ = event_tx.send(ProxyEvent::Data {
+                                stream_id,
+                                payload: bytes::Bytes::copy_from_slice(&tcp_buf[offset..end]),
+                            }).await;
+                            offset = end;
+                        }
                     }
                     Err(_) => {
                         let _ = event_tx.send(ProxyEvent::Close { stream_id }).await;
-                        if debug {
-                            eprintln!("[ostp-client] proxy CLOSE stream_id={stream_id}");
+                        if true {
+                            tracing::info!("proxy CLOSE stream_id={stream_id}");
                         }
                         break;
                     }
@@ -344,7 +864,7 @@ async fn handle_proxy_client(
                     Some(ProxyToClientMsg::Close) | Some(ProxyToClientMsg::Error(_)) | None => {
                         break;
                     }
-                    Some(ProxyToClientMsg::ConnectOk) => {} // ignored after connect phase
+                    Some(ProxyToClientMsg::ConnectOk) | Some(ProxyToClientMsg::UdpData(_, _)) => {} // ignored after connect phase
                 }
             }
         }
@@ -354,121 +874,6 @@ async fn handle_proxy_client(
     Ok(())
 }
 
-#[derive(Clone)]
-struct ExclusionMatcher {
-    domain_suffix: Vec<String>,
-    cidrs: Vec<Cidr>,
-}
-
-impl ExclusionMatcher {
-    fn new(exclusions: &ExclusionConfig) -> Self {
-        let mut cidrs = Vec::new();
-        for ip in &exclusions.ips {
-            if let Some(cidr) = parse_cidr(ip) {
-                cidrs.push(cidr);
-            }
-        }
-
-        Self {
-            domain_suffix: exclusions
-                .domains
-                .iter()
-                .map(|d| d.trim().trim_start_matches('.').to_lowercase())
-                .filter(|d| !d.is_empty())
-                .collect(),
-            cidrs,
-        }
-    }
-
-    async fn should_bypass(&self, target: &str, timeout_value: Duration) -> bool {
-        let (host, port) = match split_host_port(target) {
-            Some(v) => v,
-            None => return false,
-        };
-
-        if self.match_domain(&host) {
-            return true;
-        }
-
-        if self.cidrs.is_empty() {
-            return false;
-        }
-
-        if let Ok(ip) = host.parse::<std::net::IpAddr>() {
-            return self.match_ip(&ip);
-        }
-
-        let lookup_target = (host.clone(), port);
-        match timeout(timeout_value, tokio::net::lookup_host(lookup_target)).await {
-            Ok(Ok(addrs)) => addrs.into_iter().any(|addr| self.match_ip(&addr.ip())),
-            _ => false,
-        }
-    }
-
-    fn match_domain(&self, host: &str) -> bool {
-        if self.domain_suffix.is_empty() {
-            return false;
-        }
-        let host = host.trim_end_matches('.').to_lowercase();
-        self.domain_suffix.iter().any(|suffix| {
-            host == *suffix || host.ends_with(&format!(".{suffix}"))
-        })
-    }
-
-    fn match_ip(&self, ip: &std::net::IpAddr) -> bool {
-        self.cidrs.iter().any(|cidr| cidr.contains(ip))
-    }
-}
-
-#[derive(Clone)]
-enum Cidr {
-    V4(u32, u8),
-    V6(u128, u8),
-}
-
-impl Cidr {
-    fn contains(&self, ip: &std::net::IpAddr) -> bool {
-        match (self, ip) {
-            (Cidr::V4(net, bits), std::net::IpAddr::V4(addr)) => {
-                let mask = if *bits == 0 { 0 } else { u32::MAX << (32 - bits) };
-                let ip = u32::from_be_bytes(addr.octets());
-                (ip & mask) == (*net & mask)
-            }
-            (Cidr::V6(net, bits), std::net::IpAddr::V6(addr)) => {
-                let mask = if *bits == 0 { 0 } else { u128::MAX << (128 - bits) };
-                let ip = u128::from_be_bytes(addr.octets());
-                (ip & mask) == (*net & mask)
-            }
-            _ => false,
-        }
-    }
-}
-
-fn parse_cidr(value: &str) -> Option<Cidr> {
-    let value = value.trim();
-    if value.is_empty() {
-        return None;
-    }
-
-    if let Some((addr_str, bits_str)) = value.split_once('/') {
-        let bits: u8 = bits_str.parse().ok()?;
-        if let Ok(addr) = addr_str.parse::<std::net::IpAddr>() {
-            return match addr {
-                std::net::IpAddr::V4(v4) => Some(Cidr::V4(u32::from_be_bytes(v4.octets()), bits.min(32))),
-                std::net::IpAddr::V6(v6) => Some(Cidr::V6(u128::from_be_bytes(v6.octets()), bits.min(128))),
-            };
-        }
-    }
-
-    if let Ok(addr) = value.parse::<std::net::IpAddr>() {
-        return match addr {
-            std::net::IpAddr::V4(v4) => Some(Cidr::V4(u32::from_be_bytes(v4.octets()), 32)),
-            std::net::IpAddr::V6(v6) => Some(Cidr::V6(u128::from_be_bytes(v6.octets()), 128)),
-        };
-    }
-
-    None
-}
 
 fn split_host_port(target: &str) -> Option<(String, u16)> {
     if let Some((host, port)) = target.rsplit_once(':') {
@@ -490,14 +895,15 @@ async fn direct_connect_socks5(
     mut client: TcpStream,
     stream_id: u16,
     target: &str,
+    physical_if_index: Option<u32>,
+    physical_if_name: &Option<String>,
     close_tx: mpsc::Sender<u16>,
-    debug: bool,
+    _debug: bool,
 ) -> Result<()> {
-    if debug {
-        eprintln!("[ostp-client] proxy BYPASS stream_id={stream_id} target={target}");
+    if true {
+        tracing::info!("proxy BYPASS stream_id={stream_id} target={target}");
     }
-    let mut remote = TcpStream::connect(target).await
-        .with_context(|| format!("direct connect failed: {target}"))?;
+    let mut remote = connect_bypassing_tun(target, physical_if_index, physical_if_name).await?;
 
     client.write_all(&[0x05, 0x00, 0x00, 0x01, 0, 0, 0, 0, 0, 0]).await?;
     let _ = tokio::io::copy_bidirectional(&mut client, &mut remote).await;
@@ -511,14 +917,15 @@ async fn direct_connect_http(
     target: &str,
     method: &str,
     header_bytes: Vec<u8>,
+    physical_if_index: Option<u32>,
+    physical_if_name: &Option<String>,
     close_tx: mpsc::Sender<u16>,
-    debug: bool,
+    _debug: bool,
 ) -> Result<()> {
-    if debug {
-        eprintln!("[ostp-client] proxy BYPASS stream_id={stream_id} target={target}");
+    if true {
+        tracing::info!("proxy BYPASS stream_id={stream_id} target={target}");
     }
-    let mut remote = TcpStream::connect(target).await
-        .with_context(|| format!("direct connect failed: {target}"))?;
+    let mut remote = connect_bypassing_tun(target, physical_if_index, physical_if_name).await?;
 
     if method == "CONNECT" {
         client.write_all(b"HTTP/1.1 200 Connection Established\r\nProxy-Agent: ostp/1.0\r\n\r\n").await?;

ostp-client/src/tunnel/sni_sniff.rs

@@ -0,0 +1,73 @@
+pub fn extract_sni(data: &[u8]) -> Option<String> {
+    // Basic TLS ClientHello parser
+    // Must be at least 43 bytes to contain anything useful
+    if data.len() < 43 {
+        return None;
+    }
+
+    // TLS Record layer: Handshake (22)
+    if data[0] != 0x16 {
+        return None;
+    }
+
+    // Record layer version: 0x0301 (TLS 1.0) or 0x0303 (TLS 1.2)
+    if data[1] != 0x03 {
+        return None;
+    }
+
+    // Handshake type: ClientHello (1)
+    if data[5] != 0x01 {
+        return None;
+    }
+
+    let mut pos = 43; // Skip fixed ClientHello header
+
+    // Skip Session ID
+    if pos >= data.len() { return None; }
+    let session_id_len = data[pos] as usize;
+    pos += 1 + session_id_len;
+
+    // Skip Cipher Suites
+    if pos + 2 > data.len() { return None; }
+    let cipher_suites_len = ((data[pos] as usize) << 8) | (data[pos + 1] as usize);
+    pos += 2 + cipher_suites_len;
+
+    // Skip Compression Methods
+    if pos >= data.len() { return None; }
+    let comp_methods_len = data[pos] as usize;
+    pos += 1 + comp_methods_len;
+
+    // Extensions
+    if pos + 2 > data.len() { return None; }
+    let extensions_len = ((data[pos] as usize) << 8) | (data[pos + 1] as usize);
+    pos += 2;
+
+    let extensions_end = pos + extensions_len;
+    if extensions_end > data.len() { return None; }
+
+    while pos + 4 <= extensions_end {
+        let ext_type = ((data[pos] as usize) << 8) | (data[pos + 1] as usize);
+        let ext_len = ((data[pos + 2] as usize) << 8) | (data[pos + 3] as usize);
+        pos += 4;
+
+        if ext_type == 0x0000 { // Server Name Indication (SNI)
+            if pos + 5 <= extensions_end {
+                let _list_len = ((data[pos] as usize) << 8) | (data[pos + 1] as usize);
+                let name_type = data[pos + 2];
+                if name_type == 0 { // Hostname
+                    let name_len = ((data[pos + 3] as usize) << 8) | (data[pos + 4] as usize);
+                    if pos + 5 + name_len <= extensions_end {
+                        let sni_bytes = &data[pos + 5..pos + 5 + name_len];
+                        if let Ok(sni) = std::str::from_utf8(sni_bytes) {
+                            return Some(sni.to_string());
+                        }
+                    }
+                }
+            }
+            break;
+        }
+        pos += ext_len;
+    }
+
+    None
+}

ostp-client/src/tunnel/udp_nat.rs

@@ -0,0 +1,313 @@
+use std::collections::HashMap;
+use std::net::SocketAddr;
+use std::sync::Arc;
+use tokio::sync::{mpsc, Mutex};
+use tokio::io::{AsyncReadExt, AsyncWriteExt};
+use tokio::net::{TcpStream, UdpSocket};
+use futures::StreamExt;
+
+pub async fn run_udp_nat(
+    udp_socket: netstack_smoltcp::UdpSocket,
+    proxy_addr: String,
+    debug: bool,
+    matcher: std::sync::Arc<tokio::sync::RwLock<crate::tunnel::exclusion::ExclusionMatcher>>,
+    phys_if_index: Option<u32>,
+    phys_if_name: Option<String>,
+) {
+    let (mut rx, tx) = udp_socket.split();
+    let tx = Arc::new(Mutex::new(tx));
+    
+    // map from internal client src to a channel that sends (payload, external_dst)
+    let mut sessions: HashMap<SocketAddr, mpsc::Sender<(Vec<u8>, SocketAddr)>> = HashMap::new();
+
+    let mut cleanup_tick = tokio::time::interval(std::time::Duration::from_secs(60));
+
+    loop {
+        tokio::select! {
+            packet = rx.next() => {
+                match packet {
+                    Some((payload, src, dst)) => {
+                        if payload.is_empty() { continue; }
+
+                        if !sessions.contains_key(&src) {
+                            let (session_tx, mut session_rx) = mpsc::channel::<(Vec<u8>, SocketAddr)>(1024);
+                            sessions.insert(src, session_tx);
+
+                            let proxy_addr_clone = proxy_addr.clone();
+                            let tx_clone = tx.clone();
+                            
+                            let mut should_bypass = false;
+                            {
+                                let matcher_guard = matcher.read().await;
+                                if matcher_guard.match_ip(&dst.ip()) {
+                                    should_bypass = true;
+                                    if debug {
+                                        tracing::info!("TUN UDP BYPASS (IP match): {} → {}", src, dst);
+                                    }
+                                }
+
+                                #[cfg(target_os = "windows")]
+                                if !should_bypass {
+                                    if let Some(proc_name) = crate::tunnel::process_lookup::get_process_name_from_port_udp(src.port()) {
+                                        if debug {
+                                            tracing::debug!("TUN UDP lookup: port {} -> process {}", src.port(), proc_name);
+                                        }
+                                        if matcher_guard.match_process(&proc_name) {
+                                            should_bypass = true;
+                                            if debug {
+                                                tracing::debug!("TUN UDP BYPASS (Process match): {} ({} → {})", proc_name, src, dst);
+                                            }
+                                        }
+                                    } else {
+                                        if debug {
+                                            tracing::debug!("TUN UDP lookup: port {} -> no process found", src.port());
+                                        }
+                                    }
+                                }
+                            }
+
+                            let p_if_idx = phys_if_index;
+                            let p_if_name = phys_if_name.clone();
+
+                            tokio::spawn(async move {
+                                if should_bypass {
+                                    if debug {
+                                        tracing::info!("Starting UDP BYPASS session for {}", src);
+                                    }
+                                    let res = start_udp_bypass_session(src, p_if_idx, p_if_name, &mut session_rx, tx_clone).await;
+                                    if res.is_err() {
+                                        tracing::debug!("UDP BYPASS session for {} ended: {:?}", src, res.err());
+                                    }
+                                } else {
+                                    tracing::debug!("Starting UDP NAT session for {}", src);
+                                    let res = start_udp_session(src, proxy_addr_clone, &mut session_rx, tx_clone).await;
+                                    if res.is_err() {
+                                        tracing::debug!("UDP NAT session for {} ended: {:?}", src, res.err());
+                                    }
+                                }
+                            });
+                        }
+
+                        if let Some(sender) = sessions.get(&src) {
+                            match sender.try_send((payload, dst)) {
+                                Err(mpsc::error::TrySendError::Closed(_)) => {
+                                    sessions.remove(&src);
+                                }
+                                Err(mpsc::error::TrySendError::Full(_)) => {
+                                    // Drop packet to avoid blocking the TUN interface loop
+                                }
+                                Ok(_) => {}
+                            }
+                        }
+                    }
+                    None => break,
+                }
+            }
+            _ = cleanup_tick.tick() => {
+                sessions.retain(|_, sender| !sender.is_closed());
+            }
+        }
+    }
+}
+
+async fn start_udp_bypass_session(
+    client_src: SocketAddr,
+    phys_if_index: Option<u32>,
+    _phys_if_name: Option<String>,
+    session_rx: &mut mpsc::Receiver<(Vec<u8>, SocketAddr)>,
+    smoltcp_tx: Arc<Mutex<netstack_smoltcp::udp::WriteHalf>>,
+) -> anyhow::Result<()> {
+    let socket = match client_src {
+        SocketAddr::V4(_) => UdpSocket::bind("0.0.0.0:0").await?,
+        SocketAddr::V6(_) => UdpSocket::bind("[::]:0").await?,
+    };
+
+    #[cfg(target_os = "windows")]
+    if let Some(idx) = phys_if_index {
+        if let Err(e) = crate::tunnel::proxy::bind_socket_to_interface(&socket, client_src.is_ipv6(), idx) {
+            tracing::error!("TUN UDP BYPASS failed to bind to physical interface {}: {}", idx, e);
+        } else {
+            // Keep debug log
+        }
+    } else {
+        tracing::warn!("TUN UDP BYPASS has no physical interface index!");
+    }
+    
+    #[cfg(target_os = "linux")]
+    if let Some(ref name) = _phys_if_name {
+        let _ = crate::tunnel::proxy::bind_socket_to_interface(&socket, name);
+    }
+
+    // A single select! loop over both directions, rather than spawning a
+    // separate task for the read side, so the whole session - physical
+    // socket included - is torn down the moment this function returns
+    // (e.g. when session_rx closes). The previous spawned-task version left
+    // that task (and its Arc<UdpSocket> clone, keeping the OS socket fd
+    // alive) running forever after this function returned: nothing ever
+    // cancelled it, so every bypassed UDP flow (any excluded app/IP in TUN
+    // mode) leaked one socket + one task for the lifetime of the process.
+    use futures::SinkExt;
+    let mut buf = [0u8; 65536];
+    loop {
+        tokio::select! {
+            outbound = session_rx.recv() => {
+                match outbound {
+                    Some((payload, dst)) => { socket.send_to(&payload, dst).await?; }
+                    None => break,
+                }
+            }
+            inbound = socket.recv_from(&mut buf) => {
+                match inbound {
+                    Ok((n, peer)) => {
+                        let mut lock = smoltcp_tx.lock().await;
+                        let _ = lock.send((buf[..n].to_vec(), peer, client_src)).await;
+                    }
+                    Err(_) => break,
+                }
+            }
+        }
+    }
+
+    Ok(())
+}
+
+
+async fn start_udp_session(
+    client_src: SocketAddr,
+    proxy_addr: String,
+    session_rx: &mut mpsc::Receiver<(Vec<u8>, SocketAddr)>,
+    smoltcp_tx: Arc<Mutex<netstack_smoltcp::udp::WriteHalf>>,
+) -> anyhow::Result<()> {
+    // 1. TCP Connect to SOCKS5 proxy
+    let mut tcp = TcpStream::connect(&proxy_addr).await?;
+    
+    // Auth
+    tcp.write_all(&[5, 1, 0]).await?;
+    let mut buf = [0u8; 2];
+    tcp.read_exact(&mut buf).await?;
+    if buf[0] != 5 || buf[1] != 0 {
+        return Err(anyhow::anyhow!("socks5 auth rejected"));
+    }
+
+    // UDP ASSOCIATE to 0.0.0.0:0
+    tcp.write_all(&[5, 3, 0, 1, 0, 0, 0, 0, 0, 0]).await?;
+    let mut rep_hdr = [0u8; 4];
+    tcp.read_exact(&mut rep_hdr).await?;
+    if rep_hdr[1] != 0 {
+        return Err(anyhow::anyhow!("socks5 udp associate rejected"));
+    }
+
+    let mut relay_addr = match rep_hdr[3] {
+        1 => {
+            let mut addr_buf = [0u8; 6];
+            tcp.read_exact(&mut addr_buf).await?;
+            let ip = std::net::Ipv4Addr::new(addr_buf[0], addr_buf[1], addr_buf[2], addr_buf[3]);
+            let port = u16::from_be_bytes([addr_buf[4], addr_buf[5]]);
+            SocketAddr::new(std::net::IpAddr::V4(ip), port)
+        }
+        4 => {
+            let mut addr_buf = [0u8; 18];
+            tcp.read_exact(&mut addr_buf).await?;
+            let mut octets = [0u8; 16];
+            octets.copy_from_slice(&addr_buf[0..16]);
+            let ip = std::net::Ipv6Addr::from(octets);
+            let port = u16::from_be_bytes([addr_buf[16], addr_buf[17]]);
+            SocketAddr::new(std::net::IpAddr::V6(ip), port)
+        }
+        _ => return Err(anyhow::anyhow!("unsupported ATYP in UDP ASSOCIATE response")),
+    };
+    
+    // If proxy returned 0.0.0.0 or ::, use the proxy's IP
+    if relay_addr.ip().is_unspecified() {
+        if let Ok(proxy_sock) = proxy_addr.parse::<SocketAddr>() {
+            relay_addr.set_ip(proxy_sock.ip());
+        }
+    }
+
+    // Local SOCKS5 proxy always returns 127.0.0.1 (IPv4), so always bind IPv4
+    let udp = UdpSocket::bind("127.0.0.1:0").await?;
+
+    // CRITICAL for Android: protect this UDP socket so it goes out via the
+    // real physical interface, not back into the TUN (which would cause an
+    // infinite routing loop for DNS and all other UDP traffic).
+    #[cfg(target_os = "android")]
+    {
+        use std::os::unix::io::AsRawFd;
+        crate::bridge::protect_socket(udp.as_raw_fd());
+    }
+    
+    let mut buf = vec![0u8; 65536];
+    
+    let timeout = std::time::Duration::from_secs(300); // 5 min idle timeout
+    let mut tcp_buf = [0u8; 1];
+
+    loop {
+        tokio::select! {
+            res = tokio::time::timeout(timeout, session_rx.recv()) => {
+                match res {
+                    Ok(Some((payload, dst))) => {
+                        let mut packet = vec![0u8; 3]; // RSV, FRAG
+                        match dst.ip() {
+                            std::net::IpAddr::V4(v4) => { packet.push(1); packet.extend_from_slice(&v4.octets()); }
+                            std::net::IpAddr::V6(v6) => { packet.push(4); packet.extend_from_slice(&v6.octets()); }
+                        }
+                        packet.extend_from_slice(&dst.port().to_be_bytes());
+                        packet.extend_from_slice(&payload);
+                        tracing::debug!("udp_nat SENDING UDP ASSOCIATE payload len={} to relay_addr={} (original dst: {})", payload.len(), relay_addr, dst);
+                        let _ = udp.send_to(&packet, relay_addr).await;
+                    }
+                    Ok(None) => break,
+                    Err(_) => break, // timeout
+                }
+            }
+            res = udp.recv_from(&mut buf) => {
+                match res {
+                    Err(e) => {
+                        tracing::debug!("udp_nat recv_from error: {}", e);
+                        continue; // transient error, don't kill the session
+                    }
+                    Ok((len, _peer)) => {
+                        if len < 4 { continue; }
+                        let frag = buf[2];
+                        if frag != 0 { continue; } // fragment not supported
+                        let atyp = buf[3];
+                        let (header_len, remote_dst) = match atyp {
+                            1 => {
+                                if len < 10 { continue; }
+                                let ip = std::net::Ipv4Addr::new(buf[4], buf[5], buf[6], buf[7]);
+                                let port = u16::from_be_bytes([buf[8], buf[9]]);
+                                (10, SocketAddr::new(std::net::IpAddr::V4(ip), port))
+                            }
+                            4 => {
+                                if len < 22 { continue; }
+                                let mut octets = [0u8; 16];
+                                octets.copy_from_slice(&buf[4..20]);
+                                let ip = std::net::Ipv6Addr::from(octets);
+                                let port = u16::from_be_bytes([buf[20], buf[21]]);
+                                (22, SocketAddr::new(std::net::IpAddr::V6(ip), port))
+                            }
+                            _ => continue,
+                        };
+                        let payload = buf[header_len..len].to_vec();
+                        tracing::debug!("udp_nat RECEIVED UDP ASSOCIATE REPLY from {} for {} len={}", remote_dst, client_src, payload.len());
+                        use futures::SinkExt;
+                        if let Err(e) = smoltcp_tx.lock().await.send((payload, remote_dst, client_src)).await {
+                            tracing::error!("udp_nat failed to inject packet into smoltcp: {}", e);
+                        } else {
+                            tracing::debug!("udp_nat successfully injected packet into smoltcp from {} to {}", remote_dst, client_src);
+                        }
+                    }
+                }
+            }
+            // If TCP drops, UDP association is over
+            res = tcp.read(&mut tcp_buf) => {
+                match res {
+                    Ok(0) | Err(_) => break,
+                    Ok(_) => {}
+                }
+            }
+        }
+    }
+    
+    Ok(())
+}

ostp-client/src/tunnel/wintun_handler.rs

@@ -1,180 +0,0 @@
-use anyhow::{anyhow, Result};
-use tokio::sync::watch;
-
-#[cfg(target_os = "windows")]
-pub async fn run_wintun_tunnel(
-    config: crate::config::ClientConfig,
-    mut shutdown: watch::Receiver<bool>,
-) -> Result<()> {
-    use std::net::ToSocketAddrs;
-    use std::process::{Command, Stdio, Child};
-
-    struct WintunGuard {
-        server_ip_str: String,
-        child: Option<Child>,
-    }
-    
-    impl Drop for WintunGuard {
-        fn drop(&mut self) {
-            if let Some(mut child) = self.child.take() {
-                let _ = child.kill();
-            }
-            let cleanup_script = format!(
-                "$remote_ip = '{}'\n\
-                 Remove-NetRoute -DestinationPrefix \"$remote_ip/32\" -Confirm:$false -ErrorAction SilentlyContinue\n\
-                 Remove-NetRoute -DestinationPrefix \"1.1.1.1/32\" -Confirm:$false -ErrorAction SilentlyContinue\n\
-                 Remove-NetFirewallRule -DisplayName 'OSTP Tunnel*' -ErrorAction SilentlyContinue\n",
-                self.server_ip_str
-            );
-            let _ = Command::new("powershell").args(["-Command", &cleanup_script]).output();
-        }
-    }
-
-    let debug = config.debug;
-
-    if debug {
-        println!("[ostp-client] Initializing high-performance TUN tunnel via tun2socks...");
-    }
-
-    let exe = std::env::current_exe()?;
-    let dir = exe.parent().ok_or_else(|| anyhow!("failed to get binary directory"))?;
-    let tun2socks_exe = dir.join("tun2socks.exe");
-
-    if !tun2socks_exe.exists() {
-        return Err(anyhow!("tun2socks.exe not found in current directory! Please make sure the pre-packaged dependency is present near the executable."));
-    }
-
-    // 2. Resolve Server IP for routing table exclusion
-    let server_ip = config.ostp.server_addr.to_socket_addrs()
-        .map_err(|e| anyhow!("Failed to resolve remote server IP: {}", e))?
-        .next()
-        .map(|addr| addr.ip())
-        .ok_or_else(|| anyhow!("Could not resolve host IP for routing exclusion"))?;
-    
-    let server_ip_str = server_ip.to_string();
-
-    if debug {
-        println!("[ostp-client] Resolved remote server IP: {}", server_ip_str);
-    }
-
-    // 3. Run PowerShell script to configure system routes
-    if debug {
-        println!("[ostp-client] Injecting system routing tables and excluding remote proxy...");
-    }
-
-    let current_exe = std::env::current_exe()?.to_string_lossy().into_owned();
-
-    let setup_script = format!(
-        "$remote_ip = '{}'\n\
-         $exe_path = '{}'\n\
-         $route = Get-NetRoute -DestinationPrefix '0.0.0.0/0' | Where-Object {{ $_.InterfaceAlias -notmatch 'tun' -and $_.InterfaceAlias -notmatch 'wintun' }} | Sort-Object RouteMetric | Select-Object -First 1\n\
-         $gw = $route.NextHop\n\
-         $ifIndex = $route.InterfaceIndex\n\
-         # 1. Bypass route for the proxy server itself\n\
-         New-NetRoute -DestinationPrefix \"$remote_ip/32\" -NextHop $gw -InterfaceIndex $ifIndex -RouteMetric 1 -ErrorAction SilentlyContinue\n\
-         # 2. Bypass routes for all current Physical DNS servers to avoid UDP associate deadlocks\n\
-         $dns_ips = Get-DnsClientServerAddress -InterfaceIndex $ifIndex | Select-Object -ExpandProperty ServerAddresses\n\
-         foreach ($dns in $dns_ips) {{\n\
-             if ($dns -match '^\\d+\\.\\d+\\.\\d+\\.\\d+$') {{\n\
-                 New-NetRoute -DestinationPrefix \"$dns/32\" -NextHop $gw -InterfaceIndex $ifIndex -RouteMetric 1 -ErrorAction SilentlyContinue\n\
-             }}\n\
-         }}\n\
-         New-NetRoute -DestinationPrefix \"1.1.1.1/32\" -NextHop $gw -InterfaceIndex $ifIndex -RouteMetric 1 -ErrorAction SilentlyContinue\n\
-         # 3. Windows Firewall Rules\n\
-         New-NetFirewallRule -DisplayName 'OSTP Tunnel In' -Direction Inbound -Program $exe_path -Action Allow -Enabled True -ErrorAction SilentlyContinue\n\
-         New-NetFirewallRule -DisplayName 'OSTP Tunnel Out' -Direction Outbound -Program $exe_path -Action Allow -Enabled True -ErrorAction SilentlyContinue\n",
-        server_ip_str, current_exe
-    );
-
-    let out = Command::new("powershell")
-        .args(["-Command", &setup_script])
-        .output()?;
-    
-    if !out.status.success() && debug {
-        println!("[ostp-client] Warning: Setup routing returned: {}", String::from_utf8_lossy(&out.stderr));
-    }
-
-    // 4. Prepare and launch tun2socks.exe in the background
-    // Switch from SOCKS5 to HTTP protocol. This natively forces tun2socks NOT to attempt UDP Associate,
-    // preventing SOCKS5 command 3 unsupported errors while still tunneling 100% of global TCP traffic!
-    let proxy_url = format!("http://{}", config.local_proxy.bind_addr);
-    
-    if debug {
-        println!("[ostp-client] Spawning tun2socks daemon pointing to {}", proxy_url);
-    }
-
-    let mut child = Command::new(&tun2socks_exe)
-        .args([
-            "-device", "ostp_tun",
-            "-proxy", &proxy_url,
-            "-loglevel", if debug { "debug" } else { "error" }
-        ])
-        .current_dir(dir)
-        .stdout(if debug { Stdio::piped() } else { Stdio::null() })
-        .stderr(if debug { Stdio::piped() } else { Stdio::null() })
-        .spawn()
-        .map_err(|e| anyhow!("Failed to launch tun2socks.exe background process: {}", e))?;
-
-    let mut _guard = WintunGuard {
-        server_ip_str: server_ip_str.clone(),
-        child: None, // Will set below
-    };
-
-    // 5. Once tun2socks creates the interface, apply network settings (IP, metric)
-    tokio::time::sleep(tokio::time::Duration::from_secs(2)).await;
-
-    if debug {
-        println!("[ostp-client] Applying network configurations onto 'ostp_tun' interface...");
-    }
-
-    // We omit setting dnsservers on the TUN interface entirely. This allows Windows to natively fallback
-    // to the physical interface DNS servers, which are physically routed and work flawlessly.
-    let net_setup = "\
-        netsh interface ipv4 set address name=\"ostp_tun\" static 10.1.0.2 255.255.255.0 10.1.0.1\n\
-        netsh interface ipv4 set interface name=\"ostp_tun\" metric=5\n";
-    
-    let _ = Command::new("powershell")
-        .args(["-Command", net_setup])
-        .output()?;
-
-    println!("[client] TUN Tunnel established, internet traffic is now routing through OSTP.");
-
-    // 6. Spawn thread to keep logging tun2socks output if in debug mode
-    let mut stdout = child.stdout.take();
-    let mut stderr = child.stderr.take();
-    _guard.child = Some(child);
-
-    if debug {
-        std::thread::spawn(move || {
-            use std::io::{BufRead, BufReader};
-            if let Some(out) = stdout.take() {
-                let reader = BufReader::new(out);
-                for line in reader.lines().map_while(Result::ok) {
-                    println!("[tun2socks] {}", line);
-                }
-            }
-        });
-        std::thread::spawn(move || {
-            use std::io::{BufRead, BufReader};
-            if let Some(err) = stderr.take() {
-                let reader = BufReader::new(err);
-                for line in reader.lines().map_while(Result::ok) {
-                    println!("[tun2socks err] {}", line);
-                }
-            }
-        });
-    }
-
-    // 7. Wait for shutdown signal
-    let _ = shutdown.changed().await;
-
-    println!("[client] Deactivating TUN tunnel and restoring system network topology...");
-
-    // Drop guard runs cleanup automatically
-    drop(_guard);
-
-    println!("[client] TUN Tunnel stopped.");
-    
-    Ok(())
-}
-

ostp-client/test_udp.rs

@@ -0,0 +1 @@
+fn main() { let x: () = netstack_smoltcp::StackBuilder::default().build().unwrap(); }

ostp-core/Cargo.toml

@@ -6,13 +6,13 @@ license.workspace = true
 
 [dependencies]
 anyhow.workspace = true
-async-trait.workspace = true
 bytes.workspace = true
 chacha20poly1305.workspace = true
 rand.workspace = true
 snow.workspace = true
 thiserror.workspace = true
 tracing.workspace = true
-x25519-dalek.workspace = true
 sha2.workspace = true
 hmac.workspace = true
+x25519-dalek = { version = "2.0.1", features = ["static_secrets"] }
+hkdf = "0.12.0"

ostp-core/src/congestion.rs

@@ -0,0 +1,660 @@
+//! Congestion control for the OSTP protocol.
+//!
+//! Implements a simplified BBR-inspired algorithm that estimates bottleneck
+//! bandwidth and minimum RTT to determine the optimal sending rate.
+//! This replaces the fixed `retransmit_budget = 8` with an adaptive
+//! congestion window that responds to network conditions.
+//!
+//! RTO calculation follows RFC 6298:
+//!   SRTT = (1 - α) * SRTT + α * RTT       (α = 1/8)
+//!   RTTVAR = (1 - β) * RTTVAR + β * |SRTT - RTT|  (β = 1/4)
+//!   RTO = SRTT + 4 * RTTVAR
+//!   clamped to [RTO_MIN, RTO_MAX]
+
+use std::time::{Duration, Instant};
+
+/// Congestion control state for a single OSTP session.
+pub struct CongestionController {
+    /// Current congestion window in bytes (how much can be in-flight)
+    cwnd: u64,
+    /// Slow-start threshold in bytes
+    ssthresh: u64,
+    /// Current phase
+    phase: Phase,
+    /// Minimum RTT observed (for BBR-style bandwidth estimation)
+    min_rtt: Duration,
+    /// Smoothed RTT (RFC 6298 SRTT)
+    srtt: Duration,
+    /// RTT variance (RFC 6298 RTTVAR)
+    rttvar: Duration,
+    /// Whether we have received a first RTT sample
+    rtt_initialized: bool,
+    /// Bytes currently in flight (unacknowledged)
+    bytes_in_flight: u64,
+    /// Total bytes acknowledged (for bandwidth estimation)
+    total_acked: u64,
+    /// Last time we received an ACK
+    last_ack_time: Instant,
+    /// Number of loss events in the current window
+    loss_count: u32,
+    /// Pacing rate: bytes per second
+    pacing_rate: u64,
+    /// Token-bucket allowance for pacing, in bytes.
+    pacing_tokens: f64,
+    pacing_last_refill: Instant,
+    /// MTU estimate (used for cwnd → packet count conversion)
+    mtu: u64,
+    /// Min RTT expiry: re-probe after 10 seconds
+    min_rtt_stamp: Instant,
+    /// Loss events counted toward SLOW_START_LOSS_TOLERANCE within the
+    /// current SLOW_START_LOSS_WINDOW (see on_loss's SlowStart arm).
+    slow_start_losses: u32,
+    /// Start of the current loss-tolerance window.
+    slow_start_loss_window_start: Instant,
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+enum Phase {
+    /// Exponential growth until loss or ssthresh
+    SlowStart,
+    /// Probe bandwidth: additive increase
+    ProbeBandwidth,
+}
+
+/// Initial congestion window: 32 packets × MTU (IW10 is too conservative for modern links)
+const INITIAL_CWND_PACKETS: u64 = 32;
+/// Minimum cwnd: 2 packets
+const MIN_CWND_PACKETS: u64 = 2;
+/// Min RTT expiry window (after which we re-probe)
+const MIN_RTT_EXPIRY: Duration = Duration::from_secs(10);
+/// Minimum RTO (RFC 6298: 1s in TCP; we use 50ms since we own the protocol)
+/// Absolute ceiling on the congestion window, in packets. At a ~1200-byte MTU
+/// this is roughly 1.2 MB in flight — already far above the bandwidth-delay
+/// product of any link this protocol realistically runs over, so anything
+/// beyond it is standing queue, not throughput. The client previously allowed
+/// up to 16384 packets (~20 MB), which on a mobile uplink is minutes of buffer.
+const MAX_CWND_PACKETS: u64 = 1024;
+/// SRTT/min_rtt ratio at which slow start stops. Doubling is what fills a deep
+/// buffer fastest, so growth must end when the queue starts building rather
+/// than waiting for a loss that a deep buffer may never produce.
+const RTT_INFLATION_EXIT_SLOW_START: f64 = 2.0;
+/// SRTT/min_rtt ratio treated as a standing queue that must be actively drained.
+const RTT_INFLATION_BACKOFF: f64 = 4.0;
+/// How much pacing allowance may accumulate, expressed as time-at-rate.
+const PACING_BURST: Duration = Duration::from_millis(10);
+const RTO_MIN: Duration = Duration::from_millis(50);
+/// Maximum RTO
+const RTO_MAX: Duration = Duration::from_secs(16);
+/// Initial RTT estimate — 30 ms is reasonable for a well-connected VPN server.
+/// Will be replaced by first real measurement within milliseconds.
+const INITIAL_RTT: Duration = Duration::from_millis(30);
+
+/// Isolated packet loss during slow start (a single dropped frame from
+/// wireless noise, a brief LTE handover blip, etc.) is normal on real
+/// mobile/Wi-Fi links and does NOT mean the link is congested. The previous
+/// behavior exited slow start and halved cwnd on the very FIRST loss, which
+/// on any link with a non-zero background loss rate permanently downgrades
+/// the session from exponential growth to linear (+1 MTU/RTT) ProbeBandwidth
+/// growth within the first few RTTs - turning what should be a sub-second
+/// ramp-up into tens of seconds to minutes before throughput opens up
+/// (observed as: a trickle of KB/s, then a sudden jump once cwnd finally
+/// claws back up). Only treat loss as a real congestion signal - and pay
+/// the full slow-start-exit + halving cost - once this many losses land
+/// within SLOW_START_LOSS_WINDOW.
+const SLOW_START_LOSS_TOLERANCE: u32 = 3;
+/// Window within which SLOW_START_LOSS_TOLERANCE losses must land to count
+/// as sustained (rather than isolated) loss. Roughly a few RTTs on a
+/// well-connected link, generous on a slow one.
+const SLOW_START_LOSS_WINDOW: Duration = Duration::from_millis(500);
+
+impl CongestionController {
+    pub fn new(mtu: u64) -> Self {
+        let now = Instant::now();
+        let initial_cwnd = INITIAL_CWND_PACKETS * mtu;
+        // Initial pacing: deliver cwnd in ~2 RTTs to fill the pipe quickly
+        let initial_pacing = initial_cwnd * 1_000_000 / INITIAL_RTT.as_micros().max(1) as u64;
+        Self {
+            cwnd: initial_cwnd,
+            ssthresh: u64::MAX,
+            phase: Phase::SlowStart,
+            min_rtt: INITIAL_RTT,
+            srtt: INITIAL_RTT,
+            rttvar: INITIAL_RTT / 2,
+            rtt_initialized: false,
+            bytes_in_flight: 0,
+            total_acked: 0,
+            last_ack_time: now,
+            loss_count: 0,
+            pacing_rate: initial_pacing,
+            mtu,
+            min_rtt_stamp: now,
+            slow_start_losses: 0,
+            slow_start_loss_window_start: now,
+            pacing_tokens: (INITIAL_CWND_PACKETS * mtu) as f64,
+            pacing_last_refill: now,
+        }
+    }
+
+    /// Bytes of pacing allowance available right now, without consuming any.
+    ///
+    /// Read-only so the send path can use it as an admission check before it
+    /// commits to building a datagram.
+    pub fn pacing_available(&self) -> f64 {
+        let elapsed = self.pacing_last_refill.elapsed().as_secs_f64();
+        (self.pacing_tokens + elapsed * self.pacing_rate as f64).min(self.pacing_burst())
+    }
+
+    /// Whether at least one full-size packet may be released right now.
+    pub fn can_pace_packet(&self) -> bool {
+        self.pacing_available() >= self.mtu as f64
+    }
+
+    /// Ceiling on accumulated allowance.
+    ///
+    /// Pacing intervals here are fractions of a millisecond, so releasing
+    /// strictly one packet at a time would need a sub-millisecond timer per
+    /// packet. Instead we allow a short burst — the same trade every real
+    /// pacing implementation makes — sized so the loop's existing ~10ms wakeups
+    /// can still saturate the configured rate, with a small floor so a
+    /// cold/low estimate can never wedge sending entirely.
+    fn pacing_burst(&self) -> f64 {
+        let by_rate = self.pacing_rate as f64 * PACING_BURST.as_secs_f64();
+        by_rate.max((self.mtu * 4) as f64)
+    }
+
+    /// Refill from elapsed time and deduct `bytes`. Called on the real send
+    /// path; allowance is permitted to go negative so an oversized packet still
+    /// pays for itself rather than being released for free.
+    fn consume_pacing(&mut self, bytes: u64) {
+        let now = Instant::now();
+        let elapsed = now.duration_since(self.pacing_last_refill).as_secs_f64();
+        self.pacing_last_refill = now;
+        self.pacing_tokens =
+            (self.pacing_tokens + elapsed * self.pacing_rate as f64).min(self.pacing_burst())
+                - bytes as f64;
+    }
+
+    /// Returns the current congestion window in bytes.
+    pub fn cwnd(&self) -> u64 {
+        self.cwnd
+    }
+
+    /// Returns the current congestion window in packets.
+    pub fn cwnd_packets(&self) -> usize {
+        (self.cwnd / self.mtu).max(MIN_CWND_PACKETS) as usize
+    }
+
+    /// Returns the current pacing rate in bytes/sec.
+    pub fn pacing_rate(&self) -> u64 {
+        self.pacing_rate
+    }
+
+    /// Returns the smoothed RTT estimate (SRTT).
+    pub fn smoothed_rtt(&self) -> Duration {
+        self.srtt
+    }
+
+    /// Returns the adaptive RTO computed per RFC 6298:
+    ///   RTO = SRTT + 4 * RTTVAR, clamped to [RTO_MIN, RTO_MAX].
+    ///
+    /// This replaces the static `rto_ms` field in ProtocolMachine so that
+    /// retransmit timers automatically track changing network conditions.
+    pub fn rto(&self) -> Duration {
+        let rttvar4 = self.rttvar.saturating_mul(4);
+        let rto = self.srtt.saturating_add(rttvar4);
+        rto.clamp(RTO_MIN, RTO_MAX)
+    }
+
+    /// Returns how many bytes can still be sent.
+    pub fn available_cwnd(&self) -> u64 {
+        self.cwnd.saturating_sub(self.bytes_in_flight)
+    }
+
+    /// Returns the recommended retransmit budget per tick.
+    pub fn retransmit_budget(&self) -> usize {
+        // Allow retransmitting up to 1/4 of the cwnd in packets per tick
+        let budget = (self.cwnd_packets() / 4).max(2);
+        budget.min(64) // cap at 64 to prevent burst
+    }
+
+    /// Check whether we can send more data.
+    pub fn can_send(&self) -> bool {
+        self.bytes_in_flight < self.cwnd
+    }
+
+    /// Record that we sent `bytes` of data.
+    pub fn on_send(&mut self, bytes: u64) {
+        self.bytes_in_flight = self.bytes_in_flight.saturating_add(bytes);
+        // Charge the pacing bucket here rather than at the admission check, so
+        // every byte that actually reaches the wire is paid for exactly once —
+        // including retransmits, which are precisely what must not be allowed
+        // to bypass the rate limit and pile into an already-full queue.
+        self.consume_pacing(bytes);
+    }
+
+    /// Record that `bytes` were acknowledged but WITHOUT a usable RTT sample
+    /// (e.g. every acked frame was retransmitted, so Karn's algorithm forbids
+    /// measuring RTT from it). The window still advances; only the RTT estimator
+    /// is left untouched.
+    pub fn on_ack_no_rtt(&mut self, bytes: u64) {
+        let now = Instant::now();
+        self.bytes_in_flight = self.bytes_in_flight.saturating_sub(bytes);
+        self.total_acked = self.total_acked.saturating_add(bytes);
+        self.grow_window(bytes);
+        self.update_pacing_rate();
+        self.last_ack_time = now;
+    }
+
+    /// Record that `bytes` were acknowledged with the given RTT sample.
+    pub fn on_ack(&mut self, bytes: u64, rtt: Duration) {
+        let now = Instant::now();
+        self.bytes_in_flight = self.bytes_in_flight.saturating_sub(bytes);
+        self.total_acked = self.total_acked.saturating_add(bytes);
+
+        // Update RTT measurements
+        self.update_rtt(rtt, now);
+
+        self.grow_window(bytes);
+        self.update_pacing_rate();
+        self.last_ack_time = now;
+    }
+
+    /// Congestion-window growth shared by both ACK paths (slow start / probe).
+    fn grow_window(&mut self, bytes: u64) {
+        // ── Delay-based congestion signal ────────────────────────────────────
+        // A loss-only controller is blind on a deeply-buffered path, and mobile
+        // carrier buffers are very deep: they absorb a burst instead of dropping
+        // it, so no loss is ever signalled and cwnd keeps growing. The queue —
+        // not the link — is what grows, and the standing delay it adds shows up
+        // as RTT inflating far above the path's floor. Left unchecked this is a
+        // positive feedback loop: bigger queue -> larger RTT samples -> larger
+        // SRTT -> larger RTO -> retransmits pile on -> bigger queue, which is
+        // how a session ends up reporting multi-second (even multi-minute) RTT
+        // and stalls video until the buffer finally drains or the user
+        // reconnects. Treat sustained RTT inflation as congestion in its own
+        // right, exactly as it is.
+        let inflation = if self.rtt_initialized && !self.min_rtt.is_zero() {
+            self.srtt.as_secs_f64() / self.min_rtt.as_secs_f64()
+        } else {
+            1.0
+        };
+
+        if inflation >= RTT_INFLATION_BACKOFF {
+            // Standing queue is severe — actively drain it.
+            self.cwnd = (self.cwnd / 2).max(MIN_CWND_PACKETS * self.mtu);
+            self.ssthresh = self.cwnd;
+            self.phase = Phase::ProbeBandwidth;
+            tracing::debug!(cwnd = self.cwnd, inflation, "congestion: draining standing queue");
+            self.clamp_cwnd();
+            return;
+        }
+
+        match self.phase {
+            Phase::SlowStart => {
+                // Exponential doubling is what fills a deep buffer fastest, so
+                // leave slow start as soon as the queue starts to build rather
+                // than waiting for the loss that may never come.
+                if inflation >= RTT_INFLATION_EXIT_SLOW_START {
+                    self.ssthresh = self.cwnd;
+                    self.phase = Phase::ProbeBandwidth;
+                    tracing::debug!(cwnd = self.cwnd, inflation, "congestion: RTT inflation ended slow start");
+                    self.clamp_cwnd();
+                    return;
+                }
+                // Exponential growth: increase cwnd by acked bytes (doubles per RTT)
+                self.cwnd = self.cwnd.saturating_add(bytes);
+                if self.cwnd >= self.ssthresh {
+                    self.phase = Phase::ProbeBandwidth;
+                    tracing::debug!(cwnd = self.cwnd, "congestion: exiting slow start");
+                }
+            }
+            Phase::ProbeBandwidth => {
+                // TCP Reno Additive Increase: increase cwnd by ~1 MTU per RTT
+                self.cwnd = self.cwnd.saturating_add(bytes * self.mtu / self.cwnd.max(1));
+            }
+        }
+
+        self.clamp_cwnd();
+    }
+
+    /// Hard ceiling on the congestion window.
+    ///
+    /// Independent of any estimate: no real path this protocol runs over has a
+    /// bandwidth-delay product anywhere near this, so a window above it is
+    /// buffered queue rather than data in transit. Without it, slow start on a
+    /// buffer that never drops could grow the window into the tens of megabytes.
+    fn clamp_cwnd(&mut self) {
+        let ceiling = MAX_CWND_PACKETS.saturating_mul(self.mtu);
+        if self.cwnd > ceiling {
+            self.cwnd = ceiling;
+        }
+    }
+
+    /// Record a loss event.
+    pub fn on_loss(&mut self, bytes_lost: u64) {
+        self.bytes_in_flight = self.bytes_in_flight.saturating_sub(bytes_lost);
+        self.loss_count += 1;
+
+        match self.phase {
+            Phase::SlowStart => {
+                let now = Instant::now();
+                if now.duration_since(self.slow_start_loss_window_start) > SLOW_START_LOSS_WINDOW {
+                    // Previous window's losses have aged out - this loss starts a fresh count.
+                    self.slow_start_losses = 0;
+                    self.slow_start_loss_window_start = now;
+                }
+                self.slow_start_losses += 1;
+
+                if self.slow_start_losses >= SLOW_START_LOSS_TOLERANCE {
+                    // Sustained loss within the window: treat as real congestion.
+                    // Exit slow start, set ssthresh to half of cwnd.
+                    self.ssthresh = self.cwnd / 2;
+                    self.cwnd = self.ssthresh.max(MIN_CWND_PACKETS * self.mtu);
+                    self.phase = Phase::ProbeBandwidth;
+                    tracing::debug!(cwnd = self.cwnd, ssthresh = self.ssthresh, "congestion: sustained loss during slow start, exiting");
+                } else {
+                    // Isolated loss: likely non-congestive noise. Take a mild,
+                    // temporary haircut but keep exponential growth going -
+                    // don't throw away slow start over a single dropped frame.
+                    self.cwnd = (self.cwnd * 8 / 10).max(MIN_CWND_PACKETS * self.mtu);
+                    tracing::debug!(cwnd = self.cwnd, count = self.slow_start_losses, "congestion: isolated loss during slow start, staying in slow start");
+                }
+            }
+            Phase::ProbeBandwidth => {
+                // Multiplicative decrease: cwnd *= 0.7 (BBR-style, less aggressive than Cubic's 0.5)
+                self.cwnd = (self.cwnd * 7 / 10).max(MIN_CWND_PACKETS * self.mtu);
+                tracing::debug!(cwnd = self.cwnd, "congestion: loss, cwnd reduced");
+            }
+        }
+
+        self.update_pacing_rate();
+    }
+
+    // ── Private ──────────────────────────────────────────────────────────────
+
+    fn update_rtt(&mut self, rtt: Duration, now: Instant) {
+        // Update windowed minimum RTT (for pacing)
+        if rtt < self.min_rtt || now.duration_since(self.min_rtt_stamp) >= MIN_RTT_EXPIRY {
+            self.min_rtt = rtt;
+            self.min_rtt_stamp = now;
+        }
+
+        // Update SRTT and RTTVAR per RFC 6298
+        if !self.rtt_initialized {
+            // First measurement: initialize directly
+            self.srtt = rtt;
+            self.rttvar = rtt / 2;
+            self.rtt_initialized = true;
+        } else {
+            // RTTVAR = (3/4) * RTTVAR + (1/4) * |SRTT - R|
+            let diff = if rtt > self.srtt {
+                rtt - self.srtt
+            } else {
+                self.srtt - rtt
+            };
+            // Integer-safe: RTTVAR = RTTVAR - RTTVAR/4 + diff/4
+            self.rttvar = self.rttvar
+                .saturating_sub(self.rttvar / 4)
+                .saturating_add(diff / 4);
+
+            // SRTT = (7/8) * SRTT + (1/8) * R
+            self.srtt = self.srtt
+                .saturating_sub(self.srtt / 8)
+                .saturating_add(rtt / 8);
+        }
+
+        tracing::trace!(
+            srtt_ms = self.srtt.as_millis(),
+            rttvar_ms = self.rttvar.as_millis(),
+            rto_ms = self.rto().as_millis(),
+            "congestion: RTT updated"
+        );
+    }
+
+    fn update_pacing_rate(&mut self) {
+        // Pacing rate = cwnd / min_rtt (delivery rate target)
+        let rtt_us = self.min_rtt.as_micros().max(1) as u64;
+        self.pacing_rate = self.cwnd * 1_000_000 / rtt_us;
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_initial_state() {
+        let cc = CongestionController::new(1200);
+        assert_eq!(cc.cwnd(), 32 * 1200); // 32 * 1200
+        assert!(cc.can_send());
+        assert_eq!(cc.cwnd_packets(), 32);
+    }
+
+    #[test]
+    fn test_slow_start_growth() {
+        let mut cc = CongestionController::new(1200);
+        let initial = cc.cwnd();
+        cc.on_send(1200);
+        cc.on_ack(1200, Duration::from_millis(50));
+        assert!(cc.cwnd() > initial);
+    }
+
+    #[test]
+    fn test_loss_reduces_cwnd() {
+        let mut cc = CongestionController::new(1200);
+        let initial = cc.cwnd();
+        cc.on_loss(1200);
+        assert!(cc.cwnd() < initial);
+    }
+
+    /// The bufferbloat case: a deep buffer absorbs everything, so NOTHING is
+    /// ever lost, but the standing queue inflates RTT. A loss-only controller
+    /// grows cwnd forever here — which is how a session ends up reporting
+    /// multi-second RTT and stalling video.
+    #[test]
+    fn test_rtt_inflation_halts_growth_without_any_loss() {
+        let mut cc = CongestionController::new(1200);
+
+        // Establish a low path floor; this becomes min_rtt.
+        for _ in 0..4 {
+            cc.on_send(1200);
+            cc.on_ack(1200, Duration::from_millis(20));
+        }
+        let cwnd_before = cc.cwnd();
+
+        // Queue builds: RTT climbs far above the floor, still zero loss.
+        for _ in 0..20 {
+            cc.on_send(1200);
+            cc.on_ack(1200, Duration::from_millis(400));
+        }
+
+        assert!(
+            cc.cwnd() <= cwnd_before,
+            "cwnd kept growing while the queue was inflating RTT ({} -> {})",
+            cwnd_before,
+            cc.cwnd()
+        );
+    }
+
+    /// Pacing must actually bound the release rate: draining the bucket has to
+    /// deny the next packet. Without this the congestion window alone decides,
+    /// and a whole window leaves back-to-back.
+    #[test]
+    fn test_pacing_bucket_denies_once_drained() {
+        let mut cc = CongestionController::new(1200);
+        assert!(cc.can_pace_packet(), "a fresh controller must allow sending");
+
+        // Spend well beyond one burst allowance.
+        let burst_bytes = cc.pacing_available();
+        let mut spent = 0.0;
+        while spent <= burst_bytes + 1200.0 {
+            cc.on_send(1200);
+            spent += 1200.0;
+        }
+
+        assert!(
+            !cc.can_pace_packet(),
+            "pacing allowed unbounded sending: {} bytes still available after spending {}",
+            cc.pacing_available(),
+            spent
+        );
+    }
+
+    /// The allowance must refill over time, or sending would stall permanently
+    /// once the first burst is spent.
+    #[test]
+    fn test_pacing_bucket_refills_over_time() {
+        let mut cc = CongestionController::new(1200);
+        while cc.can_pace_packet() {
+            cc.on_send(1200);
+        }
+        assert!(!cc.can_pace_packet());
+
+        std::thread::sleep(Duration::from_millis(25));
+        assert!(
+            cc.can_pace_packet(),
+            "pacing bucket never refilled; sending would be stuck forever"
+        );
+    }
+
+    /// cwnd must never exceed the absolute ceiling, however long slow start
+    /// runs unopposed — above it the window is buffered queue, not throughput.
+    #[test]
+    fn test_cwnd_never_exceeds_absolute_ceiling() {
+        let mut cc = CongestionController::new(1200);
+        // Constant RTT: no inflation signal, so only the hard cap can stop this.
+        for _ in 0..5000 {
+            cc.on_send(1200);
+            cc.on_ack(1200, Duration::from_millis(30));
+        }
+        assert!(
+            cc.cwnd() <= MAX_CWND_PACKETS * 1200,
+            "cwnd {} exceeded the {}-packet ceiling",
+            cc.cwnd(),
+            MAX_CWND_PACKETS
+        );
+    }
+
+    #[test]
+    fn test_isolated_slow_start_loss_does_not_exit_slow_start() {
+        // A single dropped packet (wireless noise, a brief handover blip) is
+        // normal on real links and must not permanently downgrade the
+        // session from exponential to linear growth.
+        let mut cc = CongestionController::new(1200);
+        cc.on_loss(1200);
+        assert_eq!(cc.phase, Phase::SlowStart, "one isolated loss must not exit slow start");
+
+        // It should still shrink the window somewhat (not ignored entirely),
+        // just far less punishing than the sustained-congestion case.
+        let after_one = cc.cwnd();
+        assert!(after_one < INITIAL_CWND_PACKETS * 1200);
+    }
+
+    #[test]
+    fn test_sustained_slow_start_loss_exits_slow_start() {
+        // Losses landing close together (within SLOW_START_LOSS_WINDOW) are
+        // a real congestion signal and must still trigger the harsher
+        // exit-slow-start + halve response.
+        let mut cc = CongestionController::new(1200);
+        for _ in 0..SLOW_START_LOSS_TOLERANCE {
+            cc.on_loss(1200);
+        }
+        assert_eq!(cc.phase, Phase::ProbeBandwidth, "sustained loss must exit slow start");
+    }
+
+    #[test]
+    fn test_slow_start_loss_window_resets_after_expiry() {
+        // Two losses far enough apart (window expired between them) must
+        // each be treated as isolated, not accumulated toward the sustained-
+        // loss threshold.
+        let mut cc = CongestionController::new(1200);
+        cc.on_loss(1200);
+        assert_eq!(cc.phase, Phase::SlowStart);
+
+        // Simulate the window having expired by resetting its start
+        // directly (std::thread::sleep in a unit test would be flaky/slow).
+        cc.slow_start_loss_window_start = Instant::now() - SLOW_START_LOSS_WINDOW - Duration::from_millis(1);
+        cc.on_loss(1200);
+        assert_eq!(cc.phase, Phase::SlowStart, "a loss after the window expired must restart the count, not accumulate");
+        assert_eq!(cc.slow_start_losses, 1);
+    }
+
+    #[test]
+    fn test_can_send_limits() {
+        let mut cc = CongestionController::new(1200);
+        // Send until cwnd is exhausted
+        for _ in 0..32 {
+            cc.on_send(1200);
+        }
+        assert!(!cc.can_send()); // cwnd exhausted
+    }
+
+    #[test]
+    fn test_retransmit_budget() {
+        let cc = CongestionController::new(1200);
+        let budget = cc.retransmit_budget();
+        assert!(budget >= 2);
+        assert!(budget <= 64);
+    }
+
+    #[test]
+    fn test_rtt_tracking_first_sample() {
+        let mut cc = CongestionController::new(1200);
+        cc.on_send(1200);
+        cc.on_ack(1200, Duration::from_millis(25));
+        // After first sample: SRTT = 25ms, RTTVAR = 12ms
+        assert_eq!(cc.smoothed_rtt(), Duration::from_millis(25));
+    }
+
+    #[test]
+    fn test_rto_rfc6298() {
+        let mut cc = CongestionController::new(1200);
+        // After first sample with RTT=50ms: SRTT=50ms, RTTVAR=25ms, RTO=150ms
+        cc.on_send(1200);
+        cc.on_ack(1200, Duration::from_millis(50));
+        let rto = cc.rto();
+        // RTO = 50 + 4*25 = 150ms; clamped to [50ms, 16s]
+        assert!(rto >= RTO_MIN);
+        assert!(rto <= RTO_MAX);
+        assert_eq!(rto, Duration::from_millis(150));
+    }
+
+    #[test]
+    fn test_on_ack_no_rtt_grows_window_without_touching_srtt() {
+        let mut cc = CongestionController::new(1200);
+        // Establish a known SRTT with a real sample.
+        cc.on_send(1200);
+        cc.on_ack(1200, Duration::from_millis(40));
+        let srtt_before = cc.smoothed_rtt();
+        let cwnd_before = cc.cwnd();
+
+        // A Karn's-algorithm ACK (all acked frames were retransmitted): window
+        // must advance, RTT estimate must be untouched.
+        cc.on_send(1200);
+        cc.on_ack_no_rtt(1200);
+        assert!(cc.cwnd() > cwnd_before, "cwnd should still grow on a no-RTT ack");
+        assert_eq!(cc.smoothed_rtt(), srtt_before, "SRTT must not move on a no-RTT ack");
+    }
+
+    #[test]
+    fn test_rto_clamp_min() {
+        let cc = CongestionController::new(1200);
+        // Even with no RTT samples, RTO should not go below RTO_MIN
+        assert!(cc.rto() >= RTO_MIN);
+    }
+
+    #[test]
+    fn test_rto_adapts_after_multiple_samples() {
+        let mut cc = CongestionController::new(1200);
+        // Feed several consistent RTT samples
+        for _ in 0..8 {
+            cc.on_send(1200);
+            cc.on_ack(1200, Duration::from_millis(20));
+        }
+        // After convergence, RTTVAR should be small → RTO close to SRTT + small margin
+        let rto = cc.rto();
+        // Should be well below 100ms (the old hardcoded default)
+        assert!(rto < Duration::from_millis(200));
+        assert!(rto >= RTO_MIN);
+    }
+}

ostp-core/src/crypto/kex.rs

@@ -1,45 +0,0 @@
-use rand::rngs::OsRng;
-use sha2::{Digest, Sha256};
-use x25519_dalek::{EphemeralSecret, PublicKey};
-
-#[derive(Debug, Clone)]
-pub struct HybridSharedSecret {
-    pub x25519_pubkey: [u8; 32],
-    pub pq_ciphertext: Vec<u8>,
-    pub combined_secret: [u8; 32],
-}
-
-pub trait KeyExchange {
-    fn client_kex() -> HybridSharedSecret;
-}
-
-pub struct HybridKex;
-
-impl HybridKex {
-    pub fn client_offer() -> HybridSharedSecret {
-        let secret = EphemeralSecret::random_from_rng(OsRng);
-        let pubkey = PublicKey::from(&secret);
-
-        // Placeholder PQ ciphertext. Replace with ML-KEM encapsulation output.
-        let pq_ciphertext = vec![0_u8; 1088];
-        let mut hasher = Sha256::new();
-        hasher.update(pubkey.as_bytes());
-        hasher.update(&pq_ciphertext);
-        let digest = hasher.finalize();
-
-        let mut combined_secret = [0_u8; 32];
-        combined_secret.copy_from_slice(&digest[..32]);
-
-        HybridSharedSecret {
-            x25519_pubkey: *pubkey.as_bytes(),
-            pq_ciphertext,
-            combined_secret,
-        }
-    }
-}
-
-impl KeyExchange for HybridKex {
-    fn client_kex() -> HybridSharedSecret {
-        Self::client_offer()
-    }
-}

ostp-core/src/crypto/mod.rs

@@ -1,9 +1,12 @@
 pub mod aead;
-pub mod kex;
 pub mod noise;
 pub mod obfuscation;
 
+
 pub use aead::SessionCipher;
-pub use kex::{HybridSharedSecret, KeyExchange};
 pub use noise::{NoiseRole, NoiseSession};
-pub use obfuscation::{deobfuscate_packet_inplace, obfuscate_packet_inplace, derive_obfuscation_key, derive_psk};
+pub use obfuscation::{
+    deobfuscate_header_inplace, deobfuscate_packet_inplace, obfuscate_packet_inplace,
+    derive_obfuscation_key, derive_psk, derive_all_secrets, DerivedSecrets,
+    derive_junk_marker, current_junk_window, JUNK_MARKER_WINDOW_SECS,
+};

ostp-core/src/crypto/noise.rs

@@ -1,4 +1,4 @@
-use snow::{Builder, HandshakeState, TransportState};
+use snow::{Builder, HandshakeState};
 
 use crate::protocol::ProtocolError;
 
@@ -10,9 +10,15 @@ pub enum NoiseRole {
     Responder,
 }
 
-pub enum NoiseSession {
-    Handshake(HandshakeState),
-    Transport(TransportState),
+/// A Noise handshake in progress. OSTP does not use snow's transport mode: once
+/// the handshake finishes we extract the raw Split() keys (see [`raw_split`])
+/// and drive our own out-of-order AEAD (see `crypto::aead`), because the wire
+/// protocol needs explicit per-frame nonces for reordering that snow's internal
+/// nonce counter can't express.
+///
+/// [`raw_split`]: NoiseSession::raw_split
+pub struct NoiseSession {
+    handshake: Box<HandshakeState>,
 }
 
 impl NoiseSession {
@@ -36,50 +42,92 @@ impl NoiseSession {
                 .map_err(|_| ProtocolError::Crypto("noise-responder".to_string()))?,
         };
 
-        Ok(Self::Handshake(handshake))
+        Ok(Self { handshake: Box::new(handshake) })
     }
 
     pub fn write_handshake(&mut self, payload: &[u8], out: &mut [u8]) -> Result<usize, ProtocolError> {
-        match self {
-            NoiseSession::Handshake(hs) => hs
-                .write_message(payload, out)
-                .map_err(|_| ProtocolError::Crypto("noise-write".to_string())),
-            NoiseSession::Transport(_) => Err(ProtocolError::State("noise already in transport".to_string())),
-        }
+        self.handshake
+            .write_message(payload, out)
+            .map_err(|_| ProtocolError::Crypto("noise-write".to_string()))
     }
 
     pub fn read_handshake(&mut self, input: &[u8], out: &mut [u8]) -> Result<usize, ProtocolError> {
-        match self {
-            NoiseSession::Handshake(hs) => hs
-                .read_message(input, out)
-                .map_err(|_| ProtocolError::Crypto("noise-read".to_string())),
-            NoiseSession::Transport(_) => Err(ProtocolError::State("noise already in transport".to_string())),
-        }
+        self.handshake
+            .read_message(input, out)
+            .map_err(|e| ProtocolError::Crypto(format!("noise-read: {:?}", e)))
     }
 
-    pub fn handshake_hash(&self, out: &mut [u8]) -> Result<(), ProtocolError> {
-        match self {
-            NoiseSession::Handshake(hs) => {
-                let hash = hs.get_handshake_hash();
-                if out.len() != hash.len() {
-                    return Err(ProtocolError::Crypto("handshake hash length mismatch".to_string()));
-                }
-                out.copy_from_slice(hash);
-                Ok(())
-            }
-            NoiseSession::Transport(_) => Err(ProtocolError::State("noise already in transport".to_string())),
-        }
-    }
-
-    pub fn into_transport(self) -> Result<Self, ProtocolError> {
-        match self {
-            NoiseSession::Handshake(hs) => {
-                let transport = hs
-                    .into_transport_mode()
-                    .map_err(|_| ProtocolError::Crypto("noise-transport".to_string()))?;
-                Ok(NoiseSession::Transport(transport))
-            }
-            NoiseSession::Transport(_) => Ok(self),
+    /// Derive the two directional transport keys via Noise's Split().
+    ///
+    /// SECURITY: keys are taken from the final chaining key `ck` (which absorbs
+    /// the ephemeral `ee` DH result via MixKey), NOT from the handshake hash `h`
+    /// (which only absorbs public transcript data — ephemeral pubkeys and
+    /// ciphertexts — and never the DH secret). Deriving from `ck` is what gives
+    /// the session forward secrecy: an adversary who later learns the PSK still
+    /// cannot recompute these keys without the ephemeral private keys, which are
+    /// discarded after the handshake.
+    ///
+    /// Must only be called once the handshake is finished (both messages of the
+    /// NNpsk0 exchange processed); at that point `ck` is final. Returns
+    /// `(send_key, recv_key)` for the given role, matching snow's TransportState
+    /// direction mapping: split output `.0` is initiator→responder, `.1` is
+    /// responder→initiator.
+    pub fn raw_split(&mut self, role: NoiseRole) -> Result<([u8; 32], [u8; 32]), ProtocolError> {
+        if !self.handshake.is_handshake_finished() {
+            return Err(ProtocolError::State("handshake not finished at key split".to_string()));
         }
+        let (k0, k1) = self.handshake.dangerously_get_raw_split();
+        Ok(match role {
+            // Initiator sends on .0 (i→r), receives on .1 (r→i).
+            NoiseRole::Initiator => (k0, k1),
+            // Responder is the mirror image.
+            NoiseRole::Responder => (k1, k0),
+        })
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    /// Drive a full NNpsk0 handshake and confirm both sides derive matching
+    /// directional keys. This guards the .0/.1 → send/recv role mapping in
+    /// `raw_split`: if it were wrong, the two sides' send/recv keys wouldn't
+    /// cross-match and the transport channel would silently fail to decrypt.
+    #[test]
+    fn raw_split_keys_agree_across_roles() {
+        let psk = [7u8; 32];
+        let mut initiator = NoiseSession::new(NoiseRole::Initiator, &psk).unwrap();
+        let mut responder = NoiseSession::new(NoiseRole::Responder, &psk).unwrap();
+
+        // msg1: initiator -> responder
+        let mut buf1 = [0u8; 1024];
+        let n1 = initiator.write_handshake(&[], &mut buf1).unwrap();
+        let mut tmp = [0u8; 1024];
+        responder.read_handshake(&buf1[..n1], &mut tmp).unwrap();
+
+        // msg2: responder -> initiator
+        let mut buf2 = [0u8; 1024];
+        let n2 = responder.write_handshake(&[], &mut buf2).unwrap();
+        initiator.read_handshake(&buf2[..n2], &mut tmp).unwrap();
+
+        let (i_send, i_recv) = initiator.raw_split(NoiseRole::Initiator).unwrap();
+        let (r_send, r_recv) = responder.raw_split(NoiseRole::Responder).unwrap();
+
+        // What the initiator sends with, the responder must receive with.
+        assert_eq!(i_send, r_recv, "initiator send key must equal responder recv key");
+        assert_eq!(r_send, i_recv, "responder send key must equal initiator recv key");
+        // The two directions use distinct keys.
+        assert_ne!(i_send, i_recv, "the two directions must not share a key");
+    }
+
+    /// raw_split must refuse to hand out keys before the handshake is complete —
+    /// keys taken from a half-mixed chaining key would be wrong and insecure.
+    #[test]
+    fn raw_split_rejected_before_handshake_finishes() {
+        let psk = [9u8; 32];
+        let mut initiator = NoiseSession::new(NoiseRole::Initiator, &psk).unwrap();
+        // No messages exchanged yet: handshake not finished.
+        assert!(initiator.raw_split(NoiseRole::Initiator).is_err());
     }
 }

ostp-core/src/crypto/obfuscation.rs

@@ -1,88 +1,286 @@
+// =============================================================================
+// OSTP Key Derivation — Kerckhoffs's Principle
+// =============================================================================
+//
+// All protocol secrets (PSK, obfuscation key, padding parameters) are derived
+// exclusively from the access key using HKDF-SHA256. There are NO hardcoded
+// salt strings, protocol identifiers, or magic constants in this module.
+//
+// An adversary who reverse-engineers the binary sees only generic HMAC/SHA-256
+// operations with no protocol-specific strings to search for. Building a DPI
+// filter requires knowledge of the access key.
+// =============================================================================
+
 use sha2::Sha256;
 use hmac::{Hmac, Mac};
 type HmacSha256 = Hmac<Sha256>;
 
-pub fn derive_obfuscation_key(access_key: &[u8]) -> [u8; 8] {
+// ── HKDF-SHA256 (RFC 5869) ──────────────────────────────────────────────────
+// Implemented inline to avoid adding a dependency. Uses only hmac + sha2.
+
+/// HKDF-Extract: PRK = HMAC-SHA256(salt, IKM)
+fn hkdf_extract(salt: &[u8], ikm: &[u8]) -> [u8; 32] {
+    let mut mac = HmacSha256::new_from_slice(salt).expect("HMAC accepts any key length");
+    mac.update(ikm);
+    let result = mac.finalize().into_bytes();
+    let mut prk = [0u8; 32];
+    prk.copy_from_slice(&result);
+    prk
+}
+
+/// HKDF-Expand: OKM = T(1) || T(2) || ... truncated to `len` bytes.
+/// T(i) = HMAC-SHA256(PRK, T(i-1) || info || i)
+fn hkdf_expand(prk: &[u8; 32], info: &[u8], len: usize) -> Vec<u8> {
+    let mut okm = Vec::with_capacity(len);
+    let mut t = Vec::new();
+    let mut counter = 1u8;
+    while okm.len() < len {
+        let mut mac = HmacSha256::new_from_slice(prk).expect("HMAC accepts any key length");
+        mac.update(&t);
+        mac.update(info);
+        mac.update(&[counter]);
+        let block = mac.finalize().into_bytes();
+        t = block.to_vec();
+        okm.extend_from_slice(&t[..t.len().min(len - okm.len() + t.len()).min(t.len())]);
+        counter = counter.wrapping_add(1);
+    }
+    okm.truncate(len);
+    okm
+}
+
+/// Derive all protocol secrets from a single access key.
+/// Returns (obfuscation_key, psk, handshake_pad_min, handshake_pad_max).
+///
+/// The derivation uses the access key as both IKM and salt material,
+/// split into two halves. No fixed strings are used — the access key
+/// alone determines all derived values.
+#[derive(Clone)]
+pub struct DerivedSecrets {
+    pub obfuscation_key: [u8; 8],
+    pub psk: [u8; 32],
+    pub handshake_pad_min: usize,
+    pub handshake_pad_max: usize,
+}
+// NOTE: the junk marker is NOT part of DerivedSecrets — it is time-rotating and
+// derived separately per window via `derive_junk_marker` (see below), so it
+// carries no static per-user signature.
+
+/// OSTP wire protocol version. Mixed into key derivation (NOT sent on the
+/// wire) so peers running incompatible versions derive entirely different
+/// secrets and therefore cannot deobfuscate / decrypt each other's traffic.
+///
+/// This is a hard, deterministic version gate that needs NO plaintext version
+/// byte on the wire — a constant marker would defeat the project's stealth
+/// north-star ("no recognizable header"). A pre-0.4.0 client (which derived
+/// without a version) produces a different obfuscation key, so a 0.4.0 server
+/// cannot recover its handshake header and rejects it as an unauthorized probe.
+///
+/// Bump this on any wire-breaking protocol change. 0.4.0 = version 4;
+/// version 5 (0.4.x hardening) moved transport keys from the handshake hash to
+/// Noise's Split() output — a wire-breaking crypto change, so old peers must not
+/// interop (they would derive different session keys and fail decryption).
+pub const PROTOCOL_VERSION: u8 = 5;
+
+pub fn derive_all_secrets(access_key: &[u8]) -> DerivedSecrets {
+    derive_all_secrets_versioned(access_key, PROTOCOL_VERSION)
+}
+
+/// Version-parameterised derivation. `derive_all_secrets` always pins the
+/// current `PROTOCOL_VERSION`; this form exists so tests can prove that a
+/// different version yields incompatible secrets (the version gate).
+pub(crate) fn derive_all_secrets_versioned(access_key: &[u8], version: u8) -> DerivedSecrets {
+    // Split the key hash into two halves for salt/info separation.
+    // This avoids using any hardcoded strings while still providing
+    // domain separation between the derived values.
     use sha2::Digest;
-    let mut hasher = Sha256::new();
-    hasher.update(access_key);
-    let result = hasher.finalize();
-    let mut key = [0u8; 8];
-    key.copy_from_slice(&result[0..8]);
-    key
+    let key_hash = sha2::Sha256::digest(access_key);
+    let salt = &key_hash[..16];
+    let info_base = &key_hash[16..];
+
+    // Mix the protocol version into the IKM so a different version produces a
+    // completely different PRK → different obf_key / psk / padding. This is the
+    // wire-version gate: it is invisible on the wire (only the derived output,
+    // which is already indistinguishable from random, ever leaves the host).
+    let mut ikm = Vec::with_capacity(access_key.len() + 1);
+    ikm.extend_from_slice(access_key);
+    ikm.push(version);
+
+    // Extract PRK from version-tagged access key using its hash as salt
+    let prk = hkdf_extract(salt, &ikm);
+
+    // Derive obfuscation key (8 bytes) — info = key_hash[16..] || 0x01
+    let mut obf_info = info_base.to_vec();
+    obf_info.push(0x01);
+    let obf_bytes = hkdf_expand(&prk, &obf_info, 8);
+    let mut obfuscation_key = [0u8; 8];
+    obfuscation_key.copy_from_slice(&obf_bytes);
+
+    // Derive PSK (32 bytes) — info = key_hash[16..] || 0x02
+    let mut psk_info = info_base.to_vec();
+    psk_info.push(0x02);
+    let psk_bytes = hkdf_expand(&prk, &psk_info, 32);
+    let mut psk = [0u8; 32];
+    psk.copy_from_slice(&psk_bytes);
+
+    // Derive handshake padding range (2 bytes) — info = key_hash[16..] || 0x03
+    // This makes different access keys produce different handshake sizes,
+    // preventing DPI from building a universal size-based filter.
+    let mut pad_info = info_base.to_vec();
+    pad_info.push(0x03);
+    let pad_bytes = hkdf_expand(&prk, &pad_info, 2);
+    // Map to range: min ∈ [16..80], max ∈ [min+48..min+176]
+    let pad_min = 16 + (pad_bytes[0] as usize % 64);       // 16-79
+    let pad_max = pad_min + 48 + (pad_bytes[1] as usize % 128); // +48..+175
+
+    DerivedSecrets {
+        obfuscation_key,
+        psk,
+        handshake_pad_min: pad_min,
+        handshake_pad_max: pad_max,
+    }
+}
+
+/// Window length (seconds) for the rotating junk marker. The marker changes
+/// every window, so junk carries no static per-user fingerprint on the wire;
+/// the server checks the current and previous window to absorb clock skew.
+pub const JUNK_MARKER_WINDOW_SECS: u64 = 60;
+
+/// The current junk-marker time window (unix seconds / window length).
+pub fn current_junk_window() -> u64 {
+    std::time::SystemTime::now()
+        .duration_since(std::time::UNIX_EPOCH)
+        .map(|d| d.as_secs() / JUNK_MARKER_WINDOW_SECS)
+        .unwrap_or(0)
+}
+
+/// Derive the 4-byte junk marker for a given time `window`.
+///
+/// Uses the same version-gated HKDF scheme as [`derive_all_secrets`], with the
+/// window folded into the `info` (label byte `0x04`). Folding in the window
+/// makes the marker rotate: to an on-path observer the junk prefix changes every
+/// window (no fixed signature), and a captured marker is only valid for ~1
+/// window. Only a holder of the access key can compute it, so an outsider cannot
+/// forge a silently-dropped junk packet.
+pub fn derive_junk_marker(access_key: &[u8], window: u64) -> [u8; 4] {
+    derive_junk_marker_versioned(access_key, window, PROTOCOL_VERSION)
+}
+
+pub(crate) fn derive_junk_marker_versioned(access_key: &[u8], window: u64, version: u8) -> [u8; 4] {
+    use sha2::Digest;
+    let key_hash = sha2::Sha256::digest(access_key);
+    let salt = &key_hash[..16];
+    let info_base = &key_hash[16..];
+
+    let mut ikm = Vec::with_capacity(access_key.len() + 1);
+    ikm.extend_from_slice(access_key);
+    ikm.push(version);
+    let prk = hkdf_extract(salt, &ikm);
+
+    // info = key_hash[16..] || 0x04 || window(LE) — same label byte as before,
+    // now parameterised by the time window.
+    let mut info = info_base.to_vec();
+    info.push(0x04);
+    info.extend_from_slice(&window.to_le_bytes());
+    let bytes = hkdf_expand(&prk, &info, 4);
+    let mut marker = [0u8; 4];
+    marker.copy_from_slice(&bytes);
+    marker
+}
+
+// ── Legacy API (delegates to derive_all_secrets) ─────────────────────────────
+
+pub fn derive_obfuscation_key(access_key: &[u8]) -> [u8; 8] {
+    derive_all_secrets(access_key).obfuscation_key
 }
 
 pub fn derive_psk(access_key: &[u8]) -> [u8; 32] {
-    use sha2::Digest;
-    let mut hasher = Sha256::new();
-    hasher.update(access_key);
-    hasher.update(b"-ostp-psk-salt");
-    let result = hasher.finalize();
-    let mut psk = [0u8; 32];
-    psk.copy_from_slice(&result);
-    psk
+    derive_all_secrets(access_key).psk
 }
 
-/// Derives a unique 4-byte session_id mask using HMAC-SHA256(key, nonce).
-/// Because nonce is strictly monotonic, each packet gets a cryptographically
-/// independent mask — consecutive headers are indistinguishable from random noise.
-fn derive_session_mask(key: &[u8; 8], nonce: u64) -> [u8; 4] {
+// ── Wire Obfuscation ─────────────────────────────────────────────────────────
+
+/// Derives a per-packet mask from the payload following the header.
+/// Used by both data and handshake packets so every mask is unique.
+fn derive_payload_mask(key: &[u8; 8], payload: &[u8]) -> [u8; 32] {
+    let mut sample = [0u8; 32];
+    let take_len = payload.len().min(32);
+    sample[..take_len].copy_from_slice(&payload[..take_len]);
+
     let mut mac = HmacSha256::new_from_slice(key).expect("HMAC accepts any key length");
-    mac.update(&nonce.to_be_bytes());
+    mac.update(&sample);
     let result = mac.finalize().into_bytes();
-    let mut mask = [0u8; 4];
-    mask.copy_from_slice(&result[..4]);
+    let mut mask = [0u8; 32];
+    mask.copy_from_slice(&result);
     mask
 }
 
 /// Wire layout for DATA packets:
-///   [0..4]   = session_id XOR HMAC(obf_key, nonce)[0..4]   ← masked, unique per-packet
-///   [4..12]  = nonce, plaintext                             ← needed by receiver to derive mask
-///   [12..]   = AEAD ciphertext                              ← authenticates everything
+///   [0..4]   = session_id XOR mask[0..4]
+///   [4..12]  = nonce XOR mask[4..12]
+///   [12..]   = AEAD ciphertext
+///   mask = HMAC-SHA256(obf_key, ciphertext_sample[0..32])
 ///
-/// The nonce is sent in plaintext but this is intentional and safe:
-///   - It is authenticated by the AEAD tag; tampering is detected.
-///   - The session_id mask changes with every packet, breaking header correlation.
-///   - The ciphertext is fully opaque; only nonce sequence is visible.
+/// Wire layout for HANDSHAKE packets:
+///   [0..6]   = (session_id || noise_len) XOR mask[0..6]
+///   [6..]    = noise_payload || random_padding
+///   mask = HMAC-SHA256(obf_key, noise_payload_sample[0..32])
+///
+/// In both cases, the mask is derived from the payload that follows the header.
+/// Since the payload contains cryptographically random data (AEAD ciphertext
+/// or Noise ephemeral key), the mask is unique per packet, making the entire
+/// wire output indistinguishable from random noise.
 pub fn obfuscate_packet_inplace(raw: &mut [u8], key: &[u8; 8], is_handshake: bool) {
     if !is_handshake && raw.len() >= 12 {
-        // Read nonce from bytes 4..12 (plaintext on wire)
-        let nonce = u64::from_be_bytes([
-            raw[4], raw[5], raw[6], raw[7],
-            raw[8], raw[9], raw[10], raw[11],
-        ]);
-        // Mask only session_id bytes using nonce-derived mask
-        let mask = derive_session_mask(key, nonce);
-        for i in 0..4 {
+        let header_len = 12;
+        if raw.len() > header_len {
+            let ciphertext = &raw[header_len..];
+            let mask = derive_payload_mask(key, ciphertext);
+
+            for i in 0..12 {
+                raw[i] ^= mask[i];
+            }
+        }
+    } else if is_handshake && raw.len() > 6 {
+        let payload = &raw[6..];
+        let mask = derive_payload_mask(key, payload);
+
+        for i in 0..6 {
             raw[i] ^= mask[i];
         }
-        // nonce bytes 4..12 remain as-is (plaintext, authenticated by AEAD)
-    } else if raw.len() >= 4 {
-        // Handshake packets: mask session_id with a fixed handshake-phase mask
-        // u64::MAX used as sentinel to produce a distinct HMAC output from any data nonce
-        let mask = derive_session_mask(key, u64::MAX);
-        for i in 0..4 {
-            raw[i] ^= mask[i];
+    }
+}
+
+pub fn deobfuscate_header_inplace(
+    header: &mut [u8; 12],
+    ciphertext: &[u8],
+    key: &[u8; 8],
+    is_handshake: bool,
+) {
+    if !is_handshake {
+        let mask = derive_payload_mask(key, ciphertext);
+        for i in 0..12 {
+            header[i] ^= mask[i];
         }
     }
 }
 
 pub fn deobfuscate_packet_inplace(raw: &mut [u8], key: &[u8; 8], is_handshake: bool) {
     if !is_handshake && raw.len() >= 12 {
-        // Read nonce plaintext from bytes 4..12
-        let nonce = u64::from_be_bytes([
-            raw[4], raw[5], raw[6], raw[7],
-            raw[8], raw[9], raw[10], raw[11],
-        ]);
-        // Derive same mask and unmask session_id
-        let mask = derive_session_mask(key, nonce);
-        for i in 0..4 {
-            raw[i] ^= mask[i];
-        }
-    } else if raw.len() >= 4 {
-        let mask = derive_session_mask(key, u64::MAX);
-        for i in 0..4 {
+        let (header_slice, ciphertext) = raw.split_at_mut(12);
+        let mut header = [0u8; 12];
+        header.copy_from_slice(header_slice);
+        deobfuscate_header_inplace(&mut header, ciphertext, key, is_handshake);
+        header_slice.copy_from_slice(&header);
+    } else if is_handshake && raw.len() > 6 {
+        let payload = &raw[6..];
+        let mask = derive_payload_mask(key, payload);
+
+        for i in 0..6 {
             raw[i] ^= mask[i];
         }
     }
 }
+
+#[cfg(test)]
+#[path = "obfuscation_tests.rs"]
+mod obfuscation_tests;

ostp-core/src/crypto/obfuscation_tests.rs

@@ -0,0 +1,219 @@
+#[cfg(test)]
+mod tests {
+    use crate::crypto::obfuscation::*;
+
+    /// Verifies that derive_all_secrets is deterministic — same input always
+    /// produces the same output.
+    #[test]
+    fn test_derive_deterministic() {
+        let key = b"test_access_key_12345";
+        let s1 = derive_all_secrets(key);
+        let s2 = derive_all_secrets(key);
+
+        assert_eq!(s1.obfuscation_key, s2.obfuscation_key, "obf_key must be deterministic");
+        assert_eq!(s1.psk, s2.psk, "psk must be deterministic");
+        assert_eq!(s1.handshake_pad_min, s2.handshake_pad_min, "pad_min must be deterministic");
+        assert_eq!(s1.handshake_pad_max, s2.handshake_pad_max, "pad_max must be deterministic");
+    }
+
+    /// Verifies that different keys produce different secrets.
+    #[test]
+    fn test_derive_different_keys() {
+        let s1 = derive_all_secrets(b"key_alpha");
+        let s2 = derive_all_secrets(b"key_beta");
+
+        assert_ne!(s1.obfuscation_key, s2.obfuscation_key);
+        assert_ne!(s1.psk, s2.psk);
+    }
+
+    /// Verifies that the legacy API matches derive_all_secrets output.
+    #[test]
+    fn test_legacy_api_consistency() {
+        let key = b"consistency_check_key";
+        let secrets = derive_all_secrets(key);
+        assert_eq!(secrets.obfuscation_key, derive_obfuscation_key(key));
+        assert_eq!(secrets.psk, derive_psk(key));
+    }
+
+    /// Verifies handshake padding range is within valid bounds.
+    #[test]
+    fn test_padding_range_valid() {
+        for i in 0..100 {
+            let key = format!("test_key_{}", i);
+            let s = derive_all_secrets(key.as_bytes());
+            assert!(s.handshake_pad_min >= 16, "pad_min must be >= 16, got {}", s.handshake_pad_min);
+            assert!(s.handshake_pad_min < 80, "pad_min must be < 80, got {}", s.handshake_pad_min);
+            assert!(s.handshake_pad_max > s.handshake_pad_min, "pad_max must be > pad_min");
+            assert!(s.handshake_pad_max <= s.handshake_pad_min + 175,
+                "pad_max out of range: {} > {} + 175", s.handshake_pad_max, s.handshake_pad_min);
+        }
+    }
+
+    /// End-to-end test: obfuscate a handshake packet on the "client" side,
+    /// then deobfuscate on the "server" side using the same access key.
+    /// This simulates the exact flow that caused "Unauthorized probe" errors.
+    #[test]
+    fn test_handshake_obfuscation_roundtrip() {
+        let access_key = b"my_real_access_key_v2";
+        let secrets = derive_all_secrets(access_key);
+
+        // Simulate client building a handshake packet
+        let session_id: u32 = 0xDEADBEEF;
+        let fake_noise_payload = [0x42u8; 48]; // Typical Noise_NNpsk0 handshake size
+        let noise_len = fake_noise_payload.len() as u16;
+
+        let mut packet = Vec::new();
+        packet.extend_from_slice(&session_id.to_be_bytes());      // [0..4]
+        packet.extend_from_slice(&noise_len.to_be_bytes());        // [4..6]
+        packet.extend_from_slice(&fake_noise_payload);             // [6..54]
+        packet.extend_from_slice(&[0xAA; 64]);                     // padding
+
+        // Obfuscate (client side)
+        obfuscate_packet_inplace(&mut packet, &secrets.obfuscation_key, true);
+
+        // At this point, bytes [0..6] are masked and should look random
+        let masked_sid = u32::from_be_bytes([packet[0], packet[1], packet[2], packet[3]]);
+        assert_ne!(masked_sid, session_id, "session_id must be masked on wire");
+
+        // Deobfuscate (server side) — using same key
+        deobfuscate_packet_inplace(&mut packet, &secrets.obfuscation_key, true);
+
+        // Verify session_id is recovered
+        let recovered_sid = u32::from_be_bytes([packet[0], packet[1], packet[2], packet[3]]);
+        assert_eq!(recovered_sid, session_id, "session_id must be recovered after deobfuscation");
+
+        // Verify noise_len is recovered
+        let recovered_noise_len = u16::from_be_bytes([packet[4], packet[5]]);
+        assert_eq!(recovered_noise_len, noise_len, "noise_len must be recovered");
+
+        // Verify noise payload is intact
+        assert_eq!(&packet[6..6 + noise_len as usize], &fake_noise_payload,
+            "noise payload must be intact after round-trip");
+    }
+
+    /// Verifies that deobfuscating with the WRONG key does NOT recover
+    /// the session_id — this is what prevents unauthorized probes.
+    #[test]
+    fn test_wrong_key_produces_garbage() {
+        let correct_key = b"correct_key";
+        let wrong_key = b"wrong_key";
+
+        let correct_secrets = derive_all_secrets(correct_key);
+        let wrong_secrets = derive_all_secrets(wrong_key);
+
+        let session_id: u32 = 0x12345678;
+        let fake_noise = [0x55u8; 48];
+
+        let mut packet = Vec::new();
+        packet.extend_from_slice(&session_id.to_be_bytes());
+        packet.extend_from_slice(&(48u16).to_be_bytes());
+        packet.extend_from_slice(&fake_noise);
+        packet.extend_from_slice(&[0x00; 32]);
+
+        // Obfuscate with correct key
+        obfuscate_packet_inplace(&mut packet, &correct_secrets.obfuscation_key, true);
+
+        // Try to deobfuscate with WRONG key
+        let mut wrong_trial = packet.clone();
+        deobfuscate_packet_inplace(&mut wrong_trial, &wrong_secrets.obfuscation_key, true);
+        let wrong_sid = u32::from_be_bytes([wrong_trial[0], wrong_trial[1], wrong_trial[2], wrong_trial[3]]);
+
+        // Should NOT match — this is what the dispatcher checks
+        assert_ne!(wrong_sid, session_id, "wrong key must NOT recover session_id");
+
+        // Deobfuscate with correct key — must work
+        deobfuscate_packet_inplace(&mut packet, &correct_secrets.obfuscation_key, true);
+        let correct_sid = u32::from_be_bytes([packet[0], packet[1], packet[2], packet[3]]);
+        assert_eq!(correct_sid, session_id, "correct key must recover session_id");
+    }
+
+    /// §C version gate: a peer on a different PROTOCOL_VERSION derives
+    /// different secrets, so a handshake obfuscated with the OLD version's key
+    /// does NOT deobfuscate to a valid session_id under the current version.
+    /// This is exactly what makes an old (pre-0.4.0) client fail to connect to
+    /// a new server — with no plaintext version marker on the wire.
+    #[test]
+    fn test_protocol_version_gates_old_clients() {
+        let key = b"shared_access_key_across_versions";
+        let new = derive_all_secrets(key); // == derive_all_secrets_versioned(key, PROTOCOL_VERSION)
+        let old = derive_all_secrets_versioned(key, PROTOCOL_VERSION.wrapping_sub(1));
+
+        // Different protocol version → different derived secrets.
+        assert_ne!(new.obfuscation_key, old.obfuscation_key, "version must change obf_key");
+        assert_ne!(new.psk, old.psk, "version must change psk");
+
+        // Concretely: a handshake the old client obfuscated with its key does
+        // not recover a valid session_id when the new server deobfuscates it.
+        let session_id: u32 = 0x11223344;
+        let noise = [0x33u8; 48];
+        let mut pkt = Vec::new();
+        pkt.extend_from_slice(&session_id.to_be_bytes());
+        pkt.extend_from_slice(&(noise.len() as u16).to_be_bytes());
+        pkt.extend_from_slice(&noise);
+        pkt.extend_from_slice(&[0u8; 32]);
+
+        obfuscate_packet_inplace(&mut pkt, &old.obfuscation_key, true);   // old client
+        deobfuscate_packet_inplace(&mut pkt, &new.obfuscation_key, true); // new server
+        let recovered = u32::from_be_bytes([pkt[0], pkt[1], pkt[2], pkt[3]]);
+        assert_ne!(recovered, session_id, "old-version client must NOT be accepted by new server");
+    }
+
+    /// Verifies data packet obfuscation round-trip (non-handshake path).
+    #[test]
+    fn test_data_packet_obfuscation_roundtrip() {
+        let secrets = derive_all_secrets(b"data_test_key");
+
+        let session_id: u32 = 0xCAFEBABE;
+        let nonce: u64 = 42;
+        let ciphertext = [0x77u8; 64];
+
+        let mut packet = Vec::new();
+        packet.extend_from_slice(&session_id.to_be_bytes());  // [0..4]
+        packet.extend_from_slice(&nonce.to_be_bytes());        // [4..12]
+        packet.extend_from_slice(&ciphertext);                 // [12..]
+
+        obfuscate_packet_inplace(&mut packet, &secrets.obfuscation_key, false);
+
+        // Masked
+        let masked_sid = u32::from_be_bytes([packet[0], packet[1], packet[2], packet[3]]);
+        assert_ne!(masked_sid, session_id);
+
+        // Deobfuscate
+        deobfuscate_packet_inplace(&mut packet, &secrets.obfuscation_key, false);
+
+        let recovered_sid = u32::from_be_bytes([packet[0], packet[1], packet[2], packet[3]]);
+        let recovered_nonce = u64::from_be_bytes([
+            packet[4], packet[5], packet[6], packet[7],
+            packet[8], packet[9], packet[10], packet[11],
+        ]);
+
+        assert_eq!(recovered_sid, session_id);
+        assert_eq!(recovered_nonce, nonce);
+        assert_eq!(&packet[12..], &ciphertext);
+    }
+
+    /// The junk marker must: be stable within a window (client and server agree),
+    /// rotate across windows (no static on-wire fingerprint), and differ per key
+    /// (one user's marker never silently-drops on another user's flow).
+    #[test]
+    fn test_junk_marker_rotation() {
+        let key_a = b"access-key-alpha";
+        let key_b = b"access-key-bravo";
+
+        // Stable within a window.
+        assert_eq!(derive_junk_marker(key_a, 1000), derive_junk_marker(key_a, 1000));
+
+        // Rotates across adjacent windows.
+        assert_ne!(derive_junk_marker(key_a, 1000), derive_junk_marker(key_a, 1001));
+        assert_ne!(derive_junk_marker(key_a, 1000), derive_junk_marker(key_a, 999));
+
+        // Distinct per key within the same window.
+        assert_ne!(derive_junk_marker(key_a, 1000), derive_junk_marker(key_b, 1000));
+
+        // A different protocol version yields a different marker (version gate).
+        assert_ne!(
+            derive_junk_marker_versioned(key_a, 1000, PROTOCOL_VERSION),
+            derive_junk_marker_versioned(key_a, 1000, PROTOCOL_VERSION.wrapping_add(1)),
+        );
+    }
+}

ostp-core/src/framing/frame.rs

@@ -35,7 +35,6 @@ impl TryFrom<u8> for FrameKind {
 pub struct FrameHeader {
     pub version: u8,
     pub kind: FrameKind,
-    pub flags: u8,
     pub stream_id: u16,
     pub payload_len: u32,
     pub pad_len: u16,
@@ -45,8 +44,10 @@ impl FrameHeader {
     pub fn encode(&self, out: &mut BytesMut) {
         out.put_u8(self.version);
         out.put_u8(self.kind as u8);
-        out.put_u8(self.flags);
-        out.put_u8(0); // reserved
+        // Anti-DPI: reserved bytes filled with random data instead of zeros
+        // to prevent known-plaintext fingerprinting inside encrypted frames
+        let rnd: u16 = rand::random();
+        out.put_u16(rnd);
         out.put_u16(self.stream_id);
         out.put_u32(self.payload_len);
         out.put_u16(self.pad_len);
@@ -59,7 +60,7 @@ impl FrameHeader {
 
         let version = buf[0];
         let kind = FrameKind::try_from(buf[1])?;
-        let flags = buf[2];
+        // buf[2] and buf[3] are reserved
         let stream_id = u16::from_be_bytes([buf[4], buf[5]]);
         let payload_len = u32::from_be_bytes([buf[6], buf[7], buf[8], buf[9]]);
         let pad_len = u16::from_be_bytes([buf[10], buf[11]]);
@@ -67,7 +68,6 @@ impl FrameHeader {
         Ok(Self {
             version,
             kind,
-            flags,
             stream_id,
             payload_len,
             pad_len,
@@ -101,7 +101,15 @@ impl FramedPacket {
         let payload_len = header.payload_len as usize;
         let pad_len = header.pad_len as usize;
 
-        let expected = FRAME_HEADER_LEN + payload_len + pad_len;
+        // Use checked arithmetic: payload_len is a u32 from the (decrypted, but
+        // still to-be-trusted) header, and on 32-bit targets — MIPS/ARMv7
+        // routers are supported build targets — header+payload+pad can overflow
+        // usize and wrap to a small value that spuriously passes the length
+        // check, causing an out-of-range slice below.
+        let expected = FRAME_HEADER_LEN
+            .checked_add(payload_len)
+            .and_then(|v| v.checked_add(pad_len))
+            .ok_or_else(|| ProtocolError::Framing("frame length overflow".to_string()))?;
         if buf.len() < expected {
             return Err(ProtocolError::Framing("frame body truncated".to_string()));
         }

ostp-core/src/framing/padding.rs

@@ -10,15 +10,15 @@ pub enum TrafficProfile {
 impl TrafficProfile {
     pub fn target_size(&self, current: usize) -> usize {
         match self {
-            TrafficProfile::JsonRpc => align_up(current.max(220), 64).min(1408),
-            TrafficProfile::HttpsBurst => align_up(current.max(1200), 128).min(1472),
-            TrafficProfile::VideoStream => align_up(current.max(900), 188).min(1472),
+            TrafficProfile::JsonRpc => align_up(current.max(220), 64).min(1280),
+            TrafficProfile::HttpsBurst => align_up(current.max(1200), 128).min(1280),
+            TrafficProfile::VideoStream => align_up(current.max(900), 188).min(1280),
         }
     }
 }
 
 fn align_up(v: usize, align: usize) -> usize {
-    ((v + align - 1) / align) * align
+    v.div_ceil(align) * align
 }
 
 #[derive(Debug, Clone, Copy)]
@@ -45,11 +45,11 @@ impl AdaptivePadder {
     }
 
     pub fn padding_for_len(&self, payload_len: usize) -> usize {
-        match self.strategy {
+        let raw_pad = match self.strategy {
             PaddingStrategy::Fixed(target) => target.saturating_sub(payload_len),
             PaddingStrategy::Adaptive => {
                 let base_bucket = 64;
-                let bucketized = ((payload_len + base_bucket - 1) / base_bucket) * base_bucket;
+                let bucketized = payload_len.div_ceil(base_bucket) * base_bucket;
                 let mut target = bucketized.clamp(base_bucket, self.mtu_hint);
                 if target < payload_len {
                     target = payload_len;
@@ -60,7 +60,7 @@ impl AdaptivePadder {
                 let jitter = if jitter_cap == 0 {
                     0
                 } else {
-                    rand::thread_rng().gen_range(0..=jitter_cap.min(96))
+                    rand::thread_rng().gen_range(0..=jitter_cap.min(256))
                 };
 
                 (base_pad + jitter).min(self.max_pad)
@@ -69,7 +69,12 @@ impl AdaptivePadder {
                 let target = prof.target_size(payload_len);
                 target.saturating_sub(payload_len).min(self.max_pad)
             }
-        }
+        };
+
+        // Strict clamp to ensure total packet size (including overhead) never exceeds mtu_hint
+        let overhead = 38;
+        let max_allowed = self.mtu_hint.saturating_sub(payload_len).saturating_sub(overhead);
+        raw_pad.min(max_allowed)
     }
 
     pub fn build_padding(&self, payload_len: usize) -> Vec<u8> {

ostp-core/src/lib.rs

@@ -1,3 +1,4 @@
+pub mod congestion;
 pub mod crypto;
 pub mod framing;
 pub mod protocol;

ostp-core/src/relay.rs

@@ -10,6 +10,8 @@ pub enum RelayMessage {
     Error(String),
     Ping(u64),
     Pong(u64),
+    UdpAssociate,
+    UdpData(String, Vec<u8>),
 }
 
 impl RelayMessage {
@@ -23,6 +25,17 @@ impl RelayMessage {
             RelayMessage::Error(msg) => encode_with_len(6, msg.as_bytes()),
             RelayMessage::Ping(ts) => encode_with_len(7, &ts.to_be_bytes()),
             RelayMessage::Pong(ts) => encode_with_len(8, &ts.to_be_bytes()),
+            RelayMessage::UdpAssociate => vec![9],
+            RelayMessage::UdpData(addr, data) => {
+                let addr_bytes = addr.as_bytes();
+                let mut buf = Vec::with_capacity(1 + 2 + addr_bytes.len() + 2 + data.len());
+                buf.push(10);
+                buf.extend_from_slice(&(addr_bytes.len() as u16).to_be_bytes());
+                buf.extend_from_slice(addr_bytes);
+                buf.extend_from_slice(&(data.len() as u16).to_be_bytes());
+                buf.extend_from_slice(data);
+                buf
+            }
         }
     }
 
@@ -51,16 +64,34 @@ impl RelayMessage {
             7 => {
                 let payload = decode_with_len(&input[1..])?;
                 if payload.len() != 8 { return Err(anyhow!("invalid ping payload len")); }
-                let ts = u64::from_be_bytes(payload.try_into().unwrap());
+                let ts = u64::from_be_bytes(payload.try_into().map_err(|_| anyhow!("invalid ping payload size"))?);
                 Ok(RelayMessage::Ping(ts))
             }
             8 => {
                 let payload = decode_with_len(&input[1..])?;
-                if payload.len() != 8 { return Err(anyhow!("invalid pong payload len")); }
-                let ts = u64::from_be_bytes(payload.try_into().unwrap());
-                Ok(RelayMessage::Pong(ts))
+                if payload.len() != 8 {
+                    return Err(anyhow!("invalid pong payload"));
+                }
+                let mut ts = [0u8; 8];
+                ts.copy_from_slice(payload);
+                Ok(RelayMessage::Pong(u64::from_be_bytes(ts)))
             }
-            t => Err(anyhow!("unknown relay message type {t}")),
+            9 => Ok(RelayMessage::UdpAssociate),
+            10 => {
+                if input.len() < 3 { return Err(anyhow!("invalid udp data")); }
+                let addr_len = u16::from_be_bytes([input[1], input[2]]) as usize;
+                if input.len() < 3 + addr_len + 2 { return Err(anyhow!("invalid udp data")); }
+                let addr = String::from_utf8(input[3..3+addr_len].to_vec())
+                    .map_err(|_| anyhow!("invalid utf8 in udp addr"))?;
+                
+                let data_offset = 3 + addr_len;
+                let data_len = u16::from_be_bytes([input[data_offset], input[data_offset+1]]) as usize;
+                if input.len() < data_offset + 2 + data_len { return Err(anyhow!("invalid udp data")); }
+                
+                let data = input[data_offset+2..data_offset+2+data_len].to_vec();
+                Ok(RelayMessage::UdpData(addr, data))
+            }
+            _ => Err(anyhow!("unknown relay message type {}", input[0])),
         }
     }
 }
@@ -84,3 +115,83 @@ fn decode_with_len(input: &[u8]) -> Result<&[u8]> {
     }
     Ok(&input[2..2 + len])
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_connect_roundtrip() {
+        let msg = RelayMessage::Connect("example.com:443".to_string());
+        let encoded = msg.encode();
+        let decoded = RelayMessage::decode(&encoded).unwrap();
+        match decoded {
+            RelayMessage::Connect(addr) => assert_eq!(addr, "example.com:443"),
+            _ => panic!("expected Connect"),
+        }
+    }
+
+    #[test]
+    fn test_data_roundtrip() {
+        let data = vec![1, 2, 3, 4, 5];
+        let msg = RelayMessage::Data(data.clone());
+        let encoded = msg.encode();
+        let decoded = RelayMessage::decode(&encoded).unwrap();
+        match decoded {
+            RelayMessage::Data(d) => assert_eq!(d, data),
+            _ => panic!("expected Data"),
+        }
+    }
+
+    #[test]
+    fn test_simple_tags() {
+        assert_eq!(RelayMessage::KeepAlive.encode(), vec![3]);
+        assert_eq!(RelayMessage::Close.encode(), vec![4]);
+        assert_eq!(RelayMessage::ConnectOk.encode(), vec![5]);
+
+        assert!(matches!(RelayMessage::decode(&[3]).unwrap(), RelayMessage::KeepAlive));
+        assert!(matches!(RelayMessage::decode(&[4]).unwrap(), RelayMessage::Close));
+        assert!(matches!(RelayMessage::decode(&[5]).unwrap(), RelayMessage::ConnectOk));
+    }
+
+    #[test]
+    fn test_error_roundtrip() {
+        let msg = RelayMessage::Error("connection refused".to_string());
+        let encoded = msg.encode();
+        match RelayMessage::decode(&encoded).unwrap() {
+            RelayMessage::Error(e) => assert_eq!(e, "connection refused"),
+            _ => panic!("expected Error"),
+        }
+    }
+
+    #[test]
+    fn test_ping_pong_roundtrip() {
+        let ts = 1234567890u64;
+        match RelayMessage::decode(&RelayMessage::Ping(ts).encode()).unwrap() {
+            RelayMessage::Ping(t) => assert_eq!(t, ts),
+            _ => panic!("expected Ping"),
+        }
+        match RelayMessage::decode(&RelayMessage::Pong(ts).encode()).unwrap() {
+            RelayMessage::Pong(t) => assert_eq!(t, ts),
+            _ => panic!("expected Pong"),
+        }
+    }
+
+    #[test]
+    fn test_error_cases() {
+        assert!(RelayMessage::decode(&[]).is_err());
+        assert!(RelayMessage::decode(&[255]).is_err());
+        // Truncated: tag=1, len=5, only 2 bytes
+        assert!(RelayMessage::decode(&[1, 0, 5, b'a', b'b']).is_err());
+    }
+
+    #[test]
+    fn test_empty_data_roundtrip() {
+        let encoded = RelayMessage::Data(vec![]).encode();
+        match RelayMessage::decode(&encoded).unwrap() {
+            RelayMessage::Data(d) => assert!(d.is_empty()),
+            _ => panic!("expected Data"),
+        }
+    }
+}
+

ostp-flutter/.gitignore

@@ -0,0 +1,45 @@
+# Miscellaneous
+*.class
+*.log
+*.pyc
+*.swp
+.DS_Store
+.atom/
+.build/
+.buildlog/
+.history
+.svn/
+.swiftpm/
+migrate_working_dir/
+
+# IntelliJ related
+*.iml
+*.ipr
+*.iws
+.idea/
+
+# The .vscode folder contains launch configuration and tasks you configure in
+# VS Code which you may wish to be included in version control, so this line
+# is commented out by default.
+#.vscode/
+
+# Flutter/Dart/Pub related
+**/doc/api/
+**/ios/Flutter/.last_build_id
+.dart_tool/
+.flutter-plugins-dependencies
+.pub-cache/
+.pub/
+/build/
+/coverage/
+
+# Symbolication related
+app.*.symbols
+
+# Obfuscation related
+app.*.map.json
+
+# Android Studio will place build artifacts here
+/android/app/debug
+/android/app/profile
+/android/app/release

ostp-flutter/.metadata

@@ -0,0 +1,45 @@
+# This file tracks properties of this Flutter project.
+# Used by Flutter tool to assess capabilities and perform upgrades etc.
+#
+# This file should be version controlled and should not be manually edited.
+
+version:
+  revision: "db50e20168db8fee486b9abf32fc912de3bc5b6a"
+  channel: "stable"
+
+project_type: app
+
+# Tracks metadata for the flutter migrate command
+migration:
+  platforms:
+    - platform: root
+      create_revision: db50e20168db8fee486b9abf32fc912de3bc5b6a
+      base_revision: db50e20168db8fee486b9abf32fc912de3bc5b6a
+    - platform: android
+      create_revision: db50e20168db8fee486b9abf32fc912de3bc5b6a
+      base_revision: db50e20168db8fee486b9abf32fc912de3bc5b6a
+    - platform: ios
+      create_revision: db50e20168db8fee486b9abf32fc912de3bc5b6a
+      base_revision: db50e20168db8fee486b9abf32fc912de3bc5b6a
+    - platform: linux
+      create_revision: db50e20168db8fee486b9abf32fc912de3bc5b6a
+      base_revision: db50e20168db8fee486b9abf32fc912de3bc5b6a
+    - platform: macos
+      create_revision: db50e20168db8fee486b9abf32fc912de3bc5b6a
+      base_revision: db50e20168db8fee486b9abf32fc912de3bc5b6a
+    - platform: web
+      create_revision: db50e20168db8fee486b9abf32fc912de3bc5b6a
+      base_revision: db50e20168db8fee486b9abf32fc912de3bc5b6a
+    - platform: windows
+      create_revision: db50e20168db8fee486b9abf32fc912de3bc5b6a
+      base_revision: db50e20168db8fee486b9abf32fc912de3bc5b6a
+
+  # User provided section
+
+  # List of Local paths (relative to this file) that should be
+  # ignored by the migrate tool.
+  #
+  # Files that are not part of the templates will be ignored by default.
+  unmanaged_files:
+    - 'lib/main.dart'
+    - 'ios/Runner.xcodeproj/project.pbxproj'

ostp-flutter/README.md

@@ -0,0 +1,17 @@
+# ostp_client
+
+A new Flutter project.
+
+## Getting Started
+
+This project is a starting point for a Flutter application.
+
+A few resources to get you started if this is your first Flutter project:
+
+- [Learn Flutter](https://docs.flutter.dev/get-started/learn-flutter)
+- [Write your first Flutter app](https://docs.flutter.dev/get-started/codelab)
+- [Flutter learning resources](https://docs.flutter.dev/reference/learning-resources)
+
+For help getting started with Flutter development, view the
+[online documentation](https://docs.flutter.dev/), which offers tutorials,
+samples, guidance on mobile development, and a full API reference.

ostp-flutter/analysis_options.yaml

@@ -0,0 +1,28 @@
+# This file configures the analyzer, which statically analyzes Dart code to
+# check for errors, warnings, and lints.
+#
+# The issues identified by the analyzer are surfaced in the UI of Dart-enabled
+# IDEs (https://dart.dev/tools#ides-and-editors). The analyzer can also be
+# invoked from the command line by running `flutter analyze`.
+
+# The following line activates a set of recommended lints for Flutter apps,
+# packages, and plugins designed to encourage good coding practices.
+include: package:flutter_lints/flutter.yaml
+
+linter:
+  # The lint rules applied to this project can be customized in the
+  # section below to disable rules from the `package:flutter_lints/flutter.yaml`
+  # included above or to enable additional rules. A list of all available lints
+  # and their documentation is published at https://dart.dev/lints.
+  #
+  # Instead of disabling a lint rule for the entire project in the
+  # section below, it can also be suppressed for a single line of code
+  # or a specific dart file by using the `// ignore: name_of_lint` and
+  # `// ignore_for_file: name_of_lint` syntax on the line or in the file
+  # producing the lint.
+  rules:
+    # avoid_print: false  # Uncomment to disable the `avoid_print` rule
+    # prefer_single_quotes: true  # Uncomment to enable the `prefer_single_quotes` rule
+
+# Additional information about this file can be found at
+# https://dart.dev/guides/language/analysis-options

ostp-flutter/android/.gitignore

@@ -0,0 +1,14 @@
+gradle-wrapper.jar
+/.gradle
+/captures/
+/gradlew
+/gradlew.bat
+/local.properties
+GeneratedPluginRegistrant.java
+.cxx/
+
+# Remember to never publicly share your keystore.
+# See https://flutter.dev/to/reference-keystore
+key.properties
+**/*.keystore
+**/*.jks

ostp-flutter/android/app/build.gradle.kts

@@ -0,0 +1,119 @@
+import java.io.FileInputStream
+import java.util.Properties
+
+plugins {
+    id("com.android.application")
+    id("kotlin-android")
+    // The Flutter Gradle Plugin must be applied after the Android and Kotlin Gradle plugins.
+    id("dev.flutter.flutter-gradle-plugin")
+}
+
+// ── Release signing material ────────────────────────────────────────────────
+// Supplied out-of-band and never committed: either an `android/key.properties`
+// file (local release builds) or OSTP_KEYSTORE_* environment variables (CI).
+//
+// This exists because the release build used to be signed with the DEBUG
+// keystore (the stock Flutter template TODO). Android identifies an app by
+// applicationId + signing key, and refuses to update across a key change. The
+// debug keystore is auto-generated per machine, and CI runners are ephemeral,
+// so every published build carried a different random key — which is why
+// updating on top of a previous install failed with "App not installed" /
+// "unable to parse the package" and only a full uninstall+reinstall worked.
+val keystoreProperties = Properties().apply {
+    val propsFile = rootProject.file("key.properties")
+    if (propsFile.exists()) {
+        FileInputStream(propsFile).use { load(it) }
+    }
+}
+
+// Blank counts as absent. GitHub Actions substitutes an EMPTY STRING (not an
+// unset variable) for a secret that doesn't exist, so `getenv(...) ?: fallback`
+// silently kept the empty value — the elvis operator only catches null. That is
+// how an unset ANDROID_KEY_PASSWORD ended up being used as the literal key
+// password instead of falling back to the store password, producing Gradle's
+// "Get Key failed: Given final block not properly padded".
+fun signingSetting(propKey: String, envKey: String): String? =
+    (keystoreProperties.getProperty(propKey) ?: System.getenv(envKey))
+        ?.takeIf { it.isNotBlank() }
+
+val releaseStorePath: String? = signingSetting("storeFile", "OSTP_KEYSTORE_PATH")
+val hasReleaseSigning: Boolean = !releaseStorePath.isNullOrBlank()
+
+android {
+    namespace = "com.ospab.ostp_client"
+    compileSdk = flutter.compileSdkVersion
+    ndkVersion = flutter.ndkVersion
+
+    compileOptions {
+        sourceCompatibility = JavaVersion.VERSION_17
+        targetCompatibility = JavaVersion.VERSION_17
+    }
+
+    kotlinOptions {
+        jvmTarget = JavaVersion.VERSION_17.toString()
+    }
+
+    defaultConfig {
+        // TODO: Specify your own unique Application ID (https://developer.android.com/studio/build/application-id.html).
+        applicationId = "com.ospab.ostp_client"
+        // You can update the following values to match your application needs.
+        // For more information, see: https://flutter.dev/to/review-gradle-config.
+        minSdk = maxOf(flutter.minSdkVersion, 24)
+        targetSdk = flutter.targetSdkVersion
+        versionCode = flutter.versionCode
+        versionName = flutter.versionName
+        
+        ndk {
+            abiFilters += listOf("armeabi-v7a", "arm64-v8a", "x86_64")
+        }
+    }
+
+    signingConfigs {
+        create("release") {
+            if (hasReleaseSigning) {
+                val store = signingSetting("storePassword", "OSTP_KEYSTORE_PASSWORD")
+                storeFile = file(releaseStorePath!!)
+                storePassword = store
+                keyAlias = signingSetting("keyAlias", "OSTP_KEY_ALIAS")
+                // PKCS12 (the keytool default since Java 9, and what our upload
+                // keystore is) cannot hold a key password that differs from the
+                // store password — the format simply has no place to put one. So
+                // treat a missing key password as "same as the store password"
+                // instead of demanding a secret that, for this keystore, can only
+                // ever be a duplicate. An explicit value still wins, for the older
+                // JKS format where the two genuinely can differ.
+                keyPassword = signingSetting("keyPassword", "OSTP_KEY_PASSWORD") ?: store
+            }
+        }
+    }
+
+    buildTypes {
+        release {
+            // Use the real upload key when one was supplied; otherwise fall back to
+            // the debug keystore so a plain local `flutter build apk --release`
+            // still works for development. Anything PUBLISHED must take the first
+            // branch — a debug-signed build cannot be updated over, and its key is
+            // machine-local, so it also can't be reproduced later.
+            if (hasReleaseSigning) {
+                signingConfig = signingConfigs.getByName("release")
+            } else {
+                logger.warn(
+                    "OSTP: no release keystore configured (android/key.properties or " +
+                    "OSTP_KEYSTORE_PATH) - falling back to the DEBUG keystore. This APK " +
+                    "is for local use only: users cannot update over it, and the key is " +
+                    "not reproducible on another machine."
+                )
+                signingConfig = signingConfigs.getByName("debug")
+            }
+            proguardFiles(getDefaultProguardFile("proguard-android-optimize.txt"), "proguard-rules.pro")
+        }
+    }
+}
+
+flutter {
+    source = "../.."
+}
+
+dependencies {
+    implementation("androidx.core:core-ktx:1.13.1")
+}

ostp-flutter/android/app/proguard-rules.pro

@@ -0,0 +1,3 @@
+-keep class net.ostp.client.OstpClientSdk { *; }
+-keep class com.ospab.ostp_client.OstpVpnService { *; }
+-keep class com.ospab.ostp_client.MainActivity { *; }

ostp-flutter/android/app/src/debug/AndroidManifest.xml

@@ -0,0 +1,7 @@
+<manifest xmlns:android="http://schemas.android.com/apk/res/android">
+    <!-- The INTERNET permission is required for development. Specifically,
+         the Flutter tool needs it to communicate with the running application
+         to allow setting breakpoints, to provide hot reload, etc.
+    -->
+    <uses-permission android:name="android.permission.INTERNET"/>
+</manifest>

ostp-flutter/android/app/src/main/AndroidManifest.xml

@@ -0,0 +1,80 @@
+<manifest xmlns:android="http://schemas.android.com/apk/res/android">
+    <uses-permission android:name="android.permission.INTERNET"/>
+    <uses-permission android:name="android.permission.ACCESS_NETWORK_STATE"/>
+    <uses-permission android:name="android.permission.CHANGE_NETWORK_STATE"/>
+    <uses-permission android:name="android.permission.QUERY_ALL_PACKAGES"/>
+    <uses-permission android:name="android.permission.FOREGROUND_SERVICE"/>
+    <uses-permission android:name="android.permission.FOREGROUND_SERVICE_CONNECTED_DEVICE"/>
+    <uses-permission android:name="android.permission.WAKE_LOCK"/>
+    <uses-permission android:name="android.permission.POST_NOTIFICATIONS"/>
+    <application
+        android:label="ostp_client"
+        android:name="${applicationName}"
+        android:icon="@mipmap/launcher_icon"
+        android:roundIcon="@mipmap/launcher_icon_round"
+        android:extractNativeLibs="true">
+        <activity
+            android:name=".MainActivity"
+            android:exported="true"
+            android:launchMode="singleTop"
+            android:taskAffinity=""
+            android:theme="@style/LaunchTheme"
+            android:configChanges="orientation|keyboardHidden|keyboard|screenSize|smallestScreenSize|locale|layoutDirection|fontScale|screenLayout|density|uiMode"
+            android:hardwareAccelerated="true"
+            android:windowSoftInputMode="adjustResize">
+            <!-- Specifies an Android theme to apply to this Activity as soon as
+                 the Android process has started. This theme is visible to the user
+                 while the Flutter UI initializes. After that, this theme continues
+                 to determine the Window background behind the Flutter UI. -->
+            <meta-data
+              android:name="io.flutter.embedding.android.NormalTheme"
+              android:resource="@style/NormalTheme"
+              />
+            <intent-filter>
+                <action android:name="android.intent.action.MAIN"/>
+                <category android:name="android.intent.category.LAUNCHER"/>
+            </intent-filter>
+            <intent-filter>
+                <action android:name="android.service.quicksettings.action.QS_TILE_PREFERENCES"/>
+            </intent-filter>
+        </activity>
+        <!-- Don't delete the meta-data below.
+             This is used by the Flutter tool to generate GeneratedPluginRegistrant.java -->
+        <meta-data
+            android:name="flutterEmbedding"
+            android:value="2" />
+
+        <service
+            android:name=".OstpVpnService"
+            android:permission="android.permission.BIND_VPN_SERVICE"
+            android:foregroundServiceType="connectedDevice"
+            android:exported="false">
+            <intent-filter>
+                <action android:name="android.net.VpnService"/>
+            </intent-filter>
+        </service>
+
+        <!-- Quick Settings Tile -->
+        <service
+            android:name=".OstpTileService"
+            android:icon="@mipmap/launcher_icon"
+            android:label="OSTP VPN"
+            android:permission="android.permission.BIND_QUICK_SETTINGS_TILE"
+            android:exported="true">
+            <intent-filter>
+                <action android:name="android.service.quicksettings.action.QS_TILE"/>
+            </intent-filter>
+        </service>
+    </application>
+    <!-- Required to query activities that can process text, see:
+         https://developer.android.com/training/package-visibility and
+         https://developer.android.com/reference/android/content/Intent#ACTION_PROCESS_TEXT.
+
+         In particular, this is used by the Flutter engine in io.flutter.plugin.text.ProcessTextPlugin. -->
+    <queries>
+        <intent>
+            <action android:name="android.intent.action.PROCESS_TEXT"/>
+            <data android:mimeType="text/plain"/>
+        </intent>
+    </queries>
+</manifest>

ostp-flutter/android/app/src/main/kotlin/com/ospab/ostp_client/MainActivity.kt

@@ -0,0 +1,156 @@
+package com.ospab.ostp_client
+
+import android.content.Intent
+import android.net.VpnService
+import androidx.annotation.NonNull
+import io.flutter.embedding.android.FlutterActivity
+import io.flutter.embedding.engine.FlutterEngine
+import io.flutter.plugin.common.MethodChannel
+import android.content.pm.ApplicationInfo
+import android.content.pm.PackageManager
+import android.graphics.Bitmap
+import android.graphics.Canvas
+import android.util.Base64
+import java.io.ByteArrayOutputStream
+
+class MainActivity : FlutterActivity() {
+    private val CHANNEL = "com.ospab.ostp/vpn"
+    private val VPN_REQUEST_CODE = 0x0F
+    private var pendingConfigJson: String? = null
+
+    private fun getAppIconBase64(pm: PackageManager, appInfo: ApplicationInfo): String? {
+        try {
+            val drawable = pm.getApplicationIcon(appInfo)
+            val width = 96
+            val height = 96
+            val bitmap = Bitmap.createBitmap(width, height, Bitmap.Config.ARGB_8888)
+            val canvas = Canvas(bitmap)
+            drawable.setBounds(0, 0, width, height)
+            drawable.draw(canvas)
+            
+            val outputStream = ByteArrayOutputStream()
+            bitmap.compress(Bitmap.CompressFormat.PNG, 90, outputStream)
+            val byteArray = outputStream.toByteArray()
+            return Base64.encodeToString(byteArray, Base64.NO_WRAP)
+        } catch (e: Throwable) {
+            return null
+        }
+    }
+
+    override fun configureFlutterEngine(@NonNull flutterEngine: FlutterEngine) {
+        super.configureFlutterEngine(flutterEngine)
+        MethodChannel(flutterEngine.dartExecutor.binaryMessenger, CHANNEL).setMethodCallHandler { call, result ->
+            when (call.method) {
+                "saveConfig" -> {
+                    val configJson = call.argument<String>("configJson")
+                    val prefs = getSharedPreferences("OstpPrefs", android.content.Context.MODE_PRIVATE)
+                    prefs.edit().putString("latest_config_json", configJson).apply()
+                    result.success(true)
+                }
+                "startTunnel" -> {
+                    pendingConfigJson = call.argument<String>("configJson")
+                    val intent = VpnService.prepare(this)
+                    if (intent != null) {
+                        startActivityForResult(intent, VPN_REQUEST_CODE)
+                        result.success(true) 
+                    } else {
+                        startVpnService()
+                        result.success(true)
+                    }
+                }
+                "stopTunnel" -> {
+                    try {
+                        val intent = Intent(this, OstpVpnService::class.java)
+                        intent.action = "STOP"
+                        startService(intent)
+                        result.success(true)
+                    } catch (e: Throwable) {
+                        result.error("ERROR", e.message, null)
+                    }
+                }
+                "getLogs" -> {
+                    try {
+                        val logs = net.ostp.client.OstpClientSdk.getLogs()
+                        result.success(logs ?: "[]")
+                    } catch (e: Throwable) {
+                        result.error("ERROR", e.message ?: "Unknown JNI Error", null)
+                    }
+                }
+                "clearLogs" -> {
+                    try {
+                        net.ostp.client.OstpClientSdk.getLogs() // Drain
+                        result.success(true)
+                    } catch (e: Throwable) {
+                        result.error("ERROR", e.message, null)
+                    }
+                }
+                "isRunning" -> {
+                    result.success(OstpVpnService.isRunning)
+                }
+                "getMetrics" -> {
+                    try {
+                        val metrics = net.ostp.client.OstpClientSdk.getMetrics()
+                        result.success(metrics ?: "{}")
+                    } catch (e: Throwable) {
+                        // Surfaced into the in-app log viewer (not just logcat) so a
+                        // broken traffic counter is diagnosable from a user's bug
+                        // report without adb access.
+                        android.util.Log.e("MainActivity", "getMetrics failed", e)
+                        try {
+                            net.ostp.client.OstpClientSdk.addLog("getMetrics failed: ${e.javaClass.simpleName}: ${e.message}")
+                        } catch (_: Throwable) {}
+                        result.error("ERROR", e.message, null)
+                    }
+                }
+                "getInstalledApps" -> {
+                    // MethodChannel handlers run on the main/UI thread by default.
+                    // Enumerating every installed package AND decoding+re-encoding
+                    // each one's icon to PNG/base64 is expensive (100+ apps is
+                    // common) — done inline here it blocked the main thread for
+                    // 10-15s, during which Flutter couldn't render ANY frame, not
+                    // even the "loading" spinner, so the screen just appeared to
+                    // hang before jumping straight to the fully-loaded list.
+                    // Do the work on a background thread; only the final
+                    // `result.success(...)` needs to hop back onto the UI thread.
+                    val pm = packageManager
+                    Thread {
+                        try {
+                            val apps = pm.getInstalledApplications(PackageManager.GET_META_DATA)
+                            val list = apps.map { app ->
+                                val isSystem = ((app.flags and ApplicationInfo.FLAG_SYSTEM) != 0) &&
+                                               (pm.getLaunchIntentForPackage(app.packageName) == null)
+                                val iconBase64 = getAppIconBase64(pm, app)
+                                mapOf(
+                                    "name" to pm.getApplicationLabel(app).toString(),
+                                    "package" to app.packageName,
+                                    "isSystem" to isSystem,
+                                    "icon" to (iconBase64 ?: "")
+                                )
+                            }
+                            runOnUiThread { result.success(list) }
+                        } catch (e: Exception) {
+                            runOnUiThread { result.error("ERROR", e.message, null) }
+                        }
+                    }.start()
+                }
+                else -> result.notImplemented()
+            }
+        }
+    }
+
+    override fun onActivityResult(requestCode: Int, resultCode: Int, data: Intent?) {
+        if (requestCode == VPN_REQUEST_CODE && resultCode == RESULT_OK) {
+            startVpnService()
+        }
+        super.onActivityResult(requestCode, resultCode, data)
+    }
+
+    private fun startVpnService() {
+        val intent = Intent(this, OstpVpnService::class.java)
+        intent.action = "START"
+        if (pendingConfigJson != null) {
+            intent.putExtra("configJson", pendingConfigJson)
+        }
+        androidx.core.content.ContextCompat.startForegroundService(this, intent)
+    }
+}

ostp-flutter/android/app/src/main/kotlin/com/ospab/ostp_client/OstpTileService.kt

@@ -0,0 +1,106 @@
+package com.ospab.ostp_client
+
+import android.content.ComponentName
+import android.content.Context
+import android.content.Intent
+import android.os.Build
+import android.service.quicksettings.Tile
+import android.service.quicksettings.TileService
+import androidx.annotation.Keep
+import androidx.annotation.RequiresApi
+
+@Keep
+@RequiresApi(Build.VERSION_CODES.N)
+class OstpTileService : TileService() {
+
+    override fun onStartListening() {
+        super.onStartListening()
+        updateTile()
+    }
+
+    override fun onClick() {
+        super.onClick()
+        if (OstpVpnService.isRunning) {
+            // Отключить VPN
+            val stopIntent = Intent(this, OstpVpnService::class.java).apply { action = "STOP" }
+            startService(stopIntent)
+            // Обновим плитку сразу
+            qsTile?.state = Tile.STATE_INACTIVE
+            qsTile?.label = "OSTP VPN"
+            qsTile?.updateTile()
+        } else {
+            // Включить VPN напрямую
+            val prefs = getSharedPreferences("OstpPrefs", Context.MODE_PRIVATE)
+            val configJson = prefs.getString("latest_config_json", null)
+            
+            if (configJson != null) {
+                // Check if VPN consent is needed
+                val vpnIntent = android.net.VpnService.prepare(this)
+                if (vpnIntent != null) {
+                    // Consent needed, launch app
+                    val appIntent = packageManager.getLaunchIntentForPackage(packageName)?.apply {
+                        addFlags(Intent.FLAG_ACTIVITY_NEW_TASK or Intent.FLAG_ACTIVITY_SINGLE_TOP)
+                    }
+                    if (appIntent != null) {
+                        startActivityAndCollapse(appIntent)
+                    }
+                    return
+                }
+
+                val startIntent = Intent(this, OstpVpnService::class.java).apply {
+                    action = "START"
+                    putExtra("configJson", configJson)
+                }
+                if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.O) {
+                    startForegroundService(startIntent)
+                } else {
+                    startService(startIntent)
+                }
+                qsTile?.state = Tile.STATE_ACTIVE
+                qsTile?.label = "OSTP VPN"
+                qsTile?.updateTile()
+            } else {
+                // Если конфигурация еще не сохранена, открыть приложение
+                val appIntent = packageManager.getLaunchIntentForPackage(packageName)?.apply {
+                    addFlags(Intent.FLAG_ACTIVITY_NEW_TASK or Intent.FLAG_ACTIVITY_SINGLE_TOP)
+                    putExtra("tile_connect", true)
+                }
+                if (appIntent != null) {
+                    startActivityAndCollapse(appIntent)
+                }
+            }
+        }
+    }
+
+    private fun updateTile() {
+        val tile = qsTile ?: return
+        if (OstpVpnService.isRunning) {
+            tile.label = "OSTP VPN"
+            tile.state = Tile.STATE_ACTIVE
+        } else {
+            tile.label = "OSTP VPN"
+            tile.state = Tile.STATE_INACTIVE
+        }
+        tile.updateTile()
+    }
+
+    companion object {
+        /**
+         * Запрашивает обновление плитки быстрых настроек.
+         * Вызывается из OstpVpnService при изменении состояния.
+         */
+        @Keep
+        fun requestListeningState(context: Context) {
+            if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.N) {
+                try {
+                    requestListeningState(
+                        context,
+                        ComponentName(context, OstpTileService::class.java)
+                    )
+                } catch (e: Exception) {
+                    // Плитка может быть не добавлена в панель — это нормально
+                }
+            }
+        }
+    }
+}

ostp-flutter/android/app/src/main/kotlin/com/ospab/ostp_client/OstpVpnService.kt

@@ -0,0 +1,307 @@
+package com.ospab.ostp_client
+
+import android.app.Notification
+import android.app.NotificationChannel
+import android.app.NotificationManager
+import android.app.PendingIntent
+import android.content.Intent
+import android.content.pm.ServiceInfo
+import android.net.VpnService
+import android.os.Build
+import android.os.ParcelFileDescriptor
+import android.os.PowerManager
+import android.util.Log
+import net.ostp.client.OstpClientSdk
+import java.io.IOException
+import androidx.annotation.Keep
+import androidx.core.app.NotificationCompat
+import androidx.core.app.NotificationManagerCompat
+import androidx.core.app.ServiceCompat
+
+@Keep
+class OstpVpnService : VpnService() {
+
+    @Keep
+    companion object {
+        @Keep
+        var isRunning = false
+        @Keep
+        var instance: OstpVpnService? = null
+
+        private const val NOTIF_ID = 1001
+        private const val CHANNEL_ID = "ostp_vpn_channel"
+        private const val WAKE_LOCK_TAG = "ostp:vpn_wakelock"
+
+        /**
+         * Called by OstpClientSdk.notifyNetworkChanged() JNI thunk.
+         */
+        @Keep
+        @JvmStatic
+        fun onNetworkChanged() {
+            android.util.Log.d("OstpVpnService", "onNetworkChanged() signaled to Rust bridge")
+        }
+    }
+
+    private var vpnInterface: ParcelFileDescriptor? = null
+    private var wakeLock: PowerManager.WakeLock? = null
+    private var networkCallback: android.net.ConnectivityManager.NetworkCallback? = null
+
+    override fun onCreate() {
+        super.onCreate()
+        instance = this
+        createNotificationChannel()
+    }
+
+    override fun onStartCommand(intent: Intent?, flags: Int, startId: Int): Int {
+        val action = intent?.action
+        if (action == "START") {
+            val configJson = intent.getStringExtra("configJson") ?: return START_NOT_STICKY
+            // Launch foreground immediately so Android doesn't kill us
+            ServiceCompat.startForeground(this, NOTIF_ID, buildNotification(connecting = true), ServiceInfo.FOREGROUND_SERVICE_TYPE_CONNECTED_DEVICE)
+            startVpn(configJson)
+        } else if (action == "STOP") {
+            stopVpn()
+        }
+        return START_STICKY
+    }
+
+    private fun createNotificationChannel() {
+        if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.O) {
+            val channel = NotificationChannel(
+                CHANNEL_ID,
+                "OSTP VPN",
+                NotificationManager.IMPORTANCE_LOW
+            ).apply {
+                description = "OSTP VPN connection status"
+                setShowBadge(false)
+            }
+            val nm = getSystemService(NotificationManager::class.java)
+            nm.createNotificationChannel(channel)
+        }
+    }
+
+    private fun buildNotification(connecting: Boolean): Notification {
+        val stopIntent = PendingIntent.getService(
+            this,
+            0,
+            Intent(this, OstpVpnService::class.java).apply { action = "STOP" },
+            PendingIntent.FLAG_UPDATE_CURRENT or PendingIntent.FLAG_IMMUTABLE
+        )
+
+        val openIntent = PendingIntent.getActivity(
+            this,
+            1,
+            packageManager.getLaunchIntentForPackage(packageName)
+                ?.apply { addFlags(Intent.FLAG_ACTIVITY_SINGLE_TOP) },
+            PendingIntent.FLAG_UPDATE_CURRENT or PendingIntent.FLAG_IMMUTABLE
+        )
+
+        val (statusText, actionLabel) = if (connecting) {
+            Pair("Подключение...", "Отмена")
+        } else {
+            Pair("Подключено", "Отключить")
+        }
+
+        return NotificationCompat.Builder(this, CHANNEL_ID)
+            .setContentTitle("OSTP VPN")
+            .setContentText(statusText)
+            .setSmallIcon(android.R.drawable.ic_lock_lock)
+            .setOngoing(true)
+            .setShowWhen(false)
+            .setContentIntent(openIntent)
+            .addAction(android.R.drawable.ic_delete, actionLabel, stopIntent)
+            .setPriority(NotificationCompat.PRIORITY_LOW)
+            .build()
+    }
+
+    fun updateNotification(connected: Boolean) {
+        try {
+            val nm = NotificationManagerCompat.from(this)
+            nm.notify(NOTIF_ID, buildNotification(connecting = !connected))
+        } catch (e: Throwable) {
+            Log.e("OstpVpnService", "Failed to update notification", e)
+        }
+        // Refresh Quick Settings tile state
+        OstpTileService.requestListeningState(applicationContext)
+    }
+
+    private fun acquireWakeLock() {
+        if (wakeLock == null) {
+            val pm = getSystemService(POWER_SERVICE) as PowerManager
+            wakeLock = pm.newWakeLock(PowerManager.PARTIAL_WAKE_LOCK, WAKE_LOCK_TAG)
+            wakeLock?.acquire(24 * 60 * 60 * 1000L) // Max 24h
+            Log.d("OstpVpnService", "WakeLock acquired")
+        }
+    }
+
+    private fun releaseWakeLock() {
+        try {
+            wakeLock?.let {
+                if (it.isHeld) it.release()
+            }
+            wakeLock = null
+            Log.d("OstpVpnService", "WakeLock released")
+        } catch (e: Throwable) {
+            Log.e("OstpVpnService", "Error releasing WakeLock", e)
+        }
+    }
+
+    private fun registerNetworkCallback() {
+        if (networkCallback != null) return
+        try {
+            val cm = getSystemService(android.content.Context.CONNECTIVITY_SERVICE) as android.net.ConnectivityManager
+            networkCallback = object : android.net.ConnectivityManager.NetworkCallback() {
+                override fun onAvailable(network: android.net.Network) {
+                    super.onAvailable(network)
+                    OstpClientSdk.notifyNetworkChanged()
+                }
+                override fun onLost(network: android.net.Network) {
+                    super.onLost(network)
+                    OstpClientSdk.notifyNetworkChanged()
+                }
+            }
+            val request = android.net.NetworkRequest.Builder()
+                .addCapability(android.net.NetworkCapabilities.NET_CAPABILITY_INTERNET)
+                .build()
+            cm.registerNetworkCallback(request, networkCallback!!)
+        } catch (e: Throwable) {
+            Log.e("OstpVpnService", "Failed to register NetworkCallback", e)
+        }
+    }
+
+    private fun unregisterNetworkCallback() {
+        try {
+            if (networkCallback != null) {
+                val cm = getSystemService(android.content.Context.CONNECTIVITY_SERVICE) as android.net.ConnectivityManager
+                cm.unregisterNetworkCallback(networkCallback!!)
+                networkCallback = null
+            }
+        } catch (e: Throwable) {
+            Log.e("OstpVpnService", "Failed to unregister NetworkCallback", e)
+        }
+    }
+
+    private fun startVpn(configJson: String) {
+        if (vpnInterface != null) return
+
+        acquireWakeLock()
+
+        try {
+            val json = org.json.JSONObject(configJson)
+            val dnsServer = json.optString("dns_server", "1.1.1.1")
+            val localProxy = json.optJSONObject("local_proxy")?.optString("bind_addr", "127.0.0.1:1088") ?: "127.0.0.1:1088"
+
+            val builder = Builder()
+                .setSession("OSTP Tunnel")
+                .addAddress("10.1.0.2", 24)
+                .addAddress("fd00:1:fd00:1:fd00:1:fd00:1", 128)
+                .addRoute("0.0.0.0", 0)
+                .addRoute("::", 0)
+                .addDnsServer(dnsServer)
+                .setMtu(Math.max(1280, json.optJSONObject("ostp")?.optInt("mtu", 1140) ?: 1140))
+
+            // Always add fallback IPv4 DNS servers
+            try { builder.addDnsServer("1.1.1.1") } catch (e: Throwable) {}
+            try { builder.addDnsServer("8.8.8.8") } catch (e: Throwable) {}
+            // NOTE: Do NOT add IPv6 DNS servers here — Android would send DNS
+            // queries over IPv6, but our smoltcp TUN stack processes them as
+            // IPv4 only, causing all DNS to silently fail on LTE (IPv6-only networks).
+
+            if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.Q) {
+                builder.allowBypass()
+            }
+                
+            try {
+                builder.allowFamily(android.system.OsConstants.AF_INET)
+                builder.allowFamily(android.system.OsConstants.AF_INET6)
+            } catch (e: Throwable) { }
+                
+            val appRules = json.optJSONObject("app_rules")
+            val mode = appRules?.optString("mode", "bypass") ?: "bypass"
+            val packages = appRules?.optJSONArray("packages")
+            
+            if (mode == "proxy") {
+                if (packages != null) {
+                    for (i in 0 until packages.length()) {
+                        val pkg = packages.getString(i)
+                        try {
+                            builder.addAllowedApplication(pkg)
+                        } catch (e: Throwable) {
+                            Log.e("OstpVpnService", "Failed to add allowed application $pkg: $e")
+                        }
+                    }
+                }
+            } else {
+                try {
+                    builder.addDisallowedApplication(applicationContext.packageName)
+                } catch (e: Throwable) {
+                    Log.e("OstpVpnService", "Failed to disallow our own package: $e")
+                }
+                
+                if (packages != null) {
+                    for (i in 0 until packages.length()) {
+                        val pkg = packages.getString(i)
+                        try {
+                            builder.addDisallowedApplication(pkg)
+                        } catch (e: Throwable) {
+                            Log.e("OstpVpnService", "Failed to add disallowed application $pkg: $e")
+                        }
+                    }
+                }
+            }
+
+            vpnInterface = builder.establish()
+            val fd = vpnInterface?.fd ?: throw Exception("Failed to get VPN FD")
+            
+            // CRITICAL: Clear O_CLOEXEC so the child process inherits the TUN file descriptor
+            try {
+                android.system.Os.fcntlInt(vpnInterface!!.fileDescriptor, android.system.OsConstants.F_SETFD, 0)
+            } catch (e: Throwable) {
+                Log.e("OstpVpnService", "Failed to clear O_CLOEXEC", e)
+            }
+
+            val success = OstpClientSdk.startClient(configJson, fd, "", localProxy)
+            if (success) {
+                Log.i("OstpVpnService", "OSTP Rust Core started successfully")
+                isRunning = true
+                updateNotification(connected = true)
+            } else {
+                Log.e("OstpVpnService", "Failed to start OSTP Rust Core")
+                stopVpn()
+            }
+
+        } catch (e: Throwable) {
+            Log.e("OstpVpnService", "Error starting VPN", e)
+            android.os.Handler(android.os.Looper.getMainLooper()).post {
+                android.widget.Toast.makeText(applicationContext, "VPN Error: ${e.message}", android.widget.Toast.LENGTH_LONG).show()
+            }
+            stopVpn()
+        }
+        
+        registerNetworkCallback()
+    }
+
+    private fun stopVpn() {
+        isRunning = false
+        releaseWakeLock()
+
+        try {
+            OstpClientSdk.stopClient()
+            vpnInterface?.close()
+            vpnInterface = null
+        } catch (e: IOException) {
+            Log.e("OstpVpnService", "Error closing VPN interface", e)
+        }
+
+        stopForeground(true)
+        OstpTileService.requestListeningState(applicationContext)
+        unregisterNetworkCallback()
+        stopSelf()
+    }
+
+    override fun onDestroy() {
+        super.onDestroy()
+        instance = null
+        stopVpn()
+    }
+}

ostp-flutter/android/app/src/main/kotlin/net/ostp/client/OstpClientSdk.kt

@@ -0,0 +1,53 @@
+package net.ostp.client
+
+import androidx.annotation.Keep
+
+@Keep
+object OstpClientSdk {
+    init {
+        System.loadLibrary("ostp_jni")
+    }
+
+    @Keep
+    @JvmStatic
+    fun protectSocket(fd: Int): Boolean {
+        var retries = 5
+        while (retries > 0) {
+            val service = com.ospab.ostp_client.OstpVpnService.instance
+            if (service != null) {
+                val res = service.protect(fd)
+                android.util.Log.i("OstpClientSdk", "VpnService.protect(socketFd=$fd) -> success=$res")
+                return res
+            }
+            android.util.Log.w("OstpClientSdk", "VpnService instance is null! Retrying... ($retries left)")
+            Thread.sleep(200)
+            retries--
+        }
+        android.util.Log.e("OstpClientSdk", "VpnService instance is null! Cannot protect socketFd=$fd")
+        return false
+    }
+
+    @Keep
+    @JvmStatic
+    external fun startClient(configJson: String, fd: Int, t2sBinPath: String, localProxy: String): Boolean
+    
+    @Keep
+    @JvmStatic
+    external fun stopClient(): Boolean
+    
+    @Keep
+    @JvmStatic
+    external fun getMetrics(): String
+    
+    @Keep
+    @JvmStatic
+    external fun getLogs(): String
+    
+    @Keep
+    @JvmStatic
+    external fun addLog(logMsg: String)
+
+    @Keep
+    @JvmStatic
+    external fun notifyNetworkChanged()
+}

ostp-flutter/android/app/src/main/res/drawable-v21/launch_background.xml

@@ -0,0 +1,12 @@
+<?xml version="1.0" encoding="utf-8"?>
+<!-- Modify this file to customize your launch splash screen -->
+<layer-list xmlns:android="http://schemas.android.com/apk/res/android">
+    <item android:drawable="?android:colorBackground" />
+
+    <!-- You can insert your own image assets here -->
+    <!-- <item>
+        <bitmap
+            android:gravity="center"
+            android:src="@mipmap/launch_image" />
+    </item> -->
+</layer-list>

ostp-flutter/android/app/src/main/res/drawable/launch_background.xml

@@ -0,0 +1,12 @@
+<?xml version="1.0" encoding="utf-8"?>
+<!-- Modify this file to customize your launch splash screen -->
+<layer-list xmlns:android="http://schemas.android.com/apk/res/android">
+    <item android:drawable="@android:color/white" />
+
+    <!-- You can insert your own image assets here -->
+    <!-- <item>
+        <bitmap
+            android:gravity="center"
+            android:src="@mipmap/launch_image" />
+    </item> -->
+</layer-list>

ostp-flutter/android/app/src/main/res/mipmap-anydpi-v26/ic_launcher.xml

@@ -0,0 +1,5 @@
+<?xml version="1.0" encoding="utf-8"?>
+<adaptive-icon xmlns:android="http://schemas.android.com/apk/res/android">
+  <foreground android:drawable="@mipmap/ic_launcher_foreground"/>
+  <background android:drawable="@color/ic_launcher_background"/>
+</adaptive-icon>

ostp-flutter/android/app/src/main/res/mipmap-anydpi-v26/launcher_icon.xml

@@ -0,0 +1,5 @@
+<?xml version="1.0" encoding="utf-8"?>
+<adaptive-icon xmlns:android="http://schemas.android.com/apk/res/android">
+  <background android:drawable="@color/ic_launcher_background"/>
+  <foreground android:drawable="@drawable/ic_launcher_foreground"/>
+</adaptive-icon>