
A Linux CLI utility that transparently routes all system traffic through the Tor network using nftables. It enables rapid IP rotation and easy toggling of global proxy settings for privacy tasks.
Features • Requirements • Installation • Usage • How It Works • Verification • Contribute
No per-application setup needed - just sudo ttp start and every connection goes through Tor.
[!CAUTION] TTP is a tool designed to aid privacy by routing traffic through Tor. However, no tool can guarantee 100% anonymity. Your safety also depends on your behavior (e.g., using a regular browser vs. Tor Browser, signing into accounts, etc.). Always use TTP as part of a multi-layered security strategy.
[!WARNING] If you are a whistleblower or are engaging in high-risk activities, DO NOT use TTP. Instead, use officially audited and reliable tools like TailsOS or the Tor Browser directly. The authors and contributors of TTP assume no responsibility for your safety or the consequences of using this software.
Legacy transparent proxy scripts (TorGhost, Anonsurf) overwrite configuration files and build iptables rulesets that fail open: when they break, traffic leaves in cleartext. TTP is built the other way round - it fails closed, and keeps nothing on disk.
| Fail-closed by construction | An isolated inet ttp nftables table with a catch-all reject and policy drop on forwarding. On a crash, a watchdog trigger or an unclean exit, traffic is either routed through Tor or blocked - never released. |
| Nothing persists | Session state, torrc, lock file and logs live only in tmpfs (/run/ttp/, /run/tor/ttp/). A reboot leaves no residue and no stale lock. |
| No per-application setup | TCP and DNS are intercepted at the network layer. No SOCKS5 settings, no proxy environment variables, no application support required. |
| DNS without rewriting your system | A mount --bind overlay on /etc/resolv.conf rather than an edit, plus a volatile drop-in that neutralises systemd-resolved, backed by a kernel-level drop on any non-loopback resolver traffic. |
| The leak claim is measured | Every containment rule is tested in an isolated network namespace against the real generated ruleset, and each test first proves it can see a leak before asserting there is none. See Verification. |
transitions) monitors Tor, the nftables chains and the DNS overlay via a
double inotify watch that catches symlink-target swapping. It repairs once,
then applies an emergency killswitch.--bypass-user,
--bypass-group) with native nftables UID/GID matching, or run a single
command outside Tor with ttp bypass <cmd> via a cgroups v2 slice.torrc.ttp-tor.service
on non-standard ports, leaving an existing Tor instance untouched.Choose the method that best fits your needs. Native packages are strongly recommended for system stability, security, and clean uninstallation.
Installing via native packages ensures that all system dependencies (tor, nftables) and kernel-level optimizations (SELinux) are managed by your OS package manager.
Download the .deb or .rpm for the version you want from the latest release - the packages are release assets and are not checked into the repository - then install it:
sudo apt install ./transparent-tor-proxy_0.4.9_all.debsudo dnf install ./transparent-tor-proxy-0.4.9-1.noarch.rpmcd packaging && makepkg -siFor instructions on how to verify the integrity and authenticity of the release assets, see the Release Verification Guide.
If you are a developer or want to install from the repository:
git clone https://github.com/onyks-os/TransparentTorProxy.git
cd TransparentTorProxy
sudo ./scripts/install.sh
[!TIP] Why use
./install.sh?
Unlike standard Python installers, this script is "intelligent". On Red Hat-based systems, it detects if SELinux is in Enforcing mode and dynamically compiles a custom policy module (fromttp_tor_policy.te) to allow Tor to bind to the non-standard ports required by TTP (9041, 9054). This kernel-level optimization cannot be performed bypip.
For installing TTP via Python-specific package managers (pipx or pip with virtual environments), see the Alternative Installation Methods Reference.
TTP is designed to be simple and lightweight. For the complete list of CLI commands, options, exit codes, and technical specifications, refer to the External Interfaces Reference.
Most network-modifying commands require root privileges (sudo):
Start the proxy:
sudo ttp start
Stop the proxy:
sudo ttp stop
Check current session status:
ttp status
Verify Tor routing and latency:
ttp check
Request a new exit IP (rotate circuits):
sudo ttp refresh
For more advanced setups and circumvention profiles, see the Advanced Security & Usage Profiles Reference or consult the External Interfaces Reference.
To confirm that the tunnel is working correctly and no leaks are present:
Verify Tor Exit IP:
curl -s https://check.torproject.org/api/ip
Verify DNS Routing:
# Should return a valid IP via Tor's DNSPort
dig +short A check.torproject.org
DNS Leak Test (Terminal):
# This TXT query SHOULD return an EMPTY output
dig +short TXT whoami.ipv4.akahelp.net
Note: An empty output is the expected behavior under Tor. Tor's transparent resolver does not support TXT records; if this command returns your real ISP's IP, you have a DNS leak.
Web-based Verification: Always perform additional tests on dnsleaktest.com and ipleak.net.
To remove TTP completely from the system:
sudo ./scripts/uninstall.sh
TTP transparently routes all network traffic by orchestrating standard Linux kernel subsystems, system utilities, and Tor's control interfaces:
flowchart LR
App["Application"] --> Local["Local Network"]
Local --> DNS["systemd-resolved (Intercepted)"]
DNS --> NFT["nftables (inet ttp table)"]
NFT --> Tor["Tor Daemon"]
Tor --> Internet["Internet"]inet ttp nftables ruleset atomically to intercept TCP and DNS traffic, redirecting them to Tor while preventing IPv6 and DoT/DoH leaks./etc/resolv.conf with a volatile RAM-backed configuration via a kernel-level bind-mount to ensure DNS calls are resolved by Tor.For a detailed walkthrough of the execution flows, system hooks, security boundaries, and modular components, please refer to the:
TTP is designed to always restore your network, even in edge cases:
| Scenario | What happens |
|---|---|
ttp stop | Zero-leak cleanup: applies teardown lockdown, gracefully shuts down Tor, executes active socket slaughter, waits 1.5s, flushes connection tracking, restores firewall and DNS (via table flush and delete), and deletes the lock file |
Ctrl+C / kill | Signal handler catches SIGINT/SIGTERM and runs normal cleanup before exit |
kill -9 / Power Outage | Next ttp start detects the orphaned lock file, clears any stale mount stacks, and auto-restores |
| Manual emergency | Run sudo ./scripts/restore-network.sh to flush all nftables rules, reset DNS, and delete the lock file |
[!WARNING]
- Tor Browser: Applications using an explicit SOCKS5 proxy will create a double Tor hop. Use a regular browser instead while TTP is active.
- DNS-over-HTTPS (DoH): Normal browsers (Firefox, Chrome, Brave, Edge) may use DoH, bypassing system DNS. TTP mitigates DoH via a 3-layer defense: (1) all outbound TCP traffic (including DoH) is redirected to Tor TransPort; (2) common DoH canary domains are mapped to
0.0.0.0intorrc; (3) public DoH IP resolvers are blocked on TCP/UDP port 443 (blocking HTTP/3 QUIC DoH). For maximum security, disable DoH / "Secure DNS" in your browser settings.- IPv6: Fully supported when available. TTP dynamically detects IPv6 loopback and routes IPv6 traffic through Tor. If the host lacks IPv6 loopback support OR if the
--no-ipv6option is passed, TTP drops all outgoing IPv6 traffic to prevent leaks.- Exit IP variation: Different connections may show different exit IPs due to Tor stream isolation.
For a full breakdown of residual risks, architectural trust boundaries, and the STRIDE threat model, see:
TTP uses a Makefile to automate and standardize the testing pipeline. This ensures that every change is verified against unit and integration tests before being committed.
[!IMPORTANT] Always run
make verifybefore pushing code. If this command fails, the code is NOT ready for production.
| Command | Goal |
|---|---|
make test | Runs fast Unit Tests locally (no root needed, fully mocked). |
make integration-debian | Runs full system tests inside a privileged Docker container (Debian). |
make integration-all | Runs integration tests for all supported distros (Debian, Fedora, Arch). |
make verify | Runs Unit Tests + All Integration Tests. |
make build | Generates native .deb and .rpm packages. |
make clean | Removes all build artifacts, caches, and temp files. |
TTP's zero-leak claim is measured, not asserted. The
Network Sandbox Engine builds
an isolated network namespace, loads TTP's real generated ruleset into it,
generates the traffic a leak would consist of, and watches the boundary veth
interface with a Scapy sniffer.
Every containment test runs twice. assert no leaks is also true when the
sniffer never started, when the interface name is wrong, or when the traffic
never left the process, so each test first runs the same stimulus with the
ruleset flushed and requires the packet to be seen. Only then does it assert
that TTP's ruleset stops it. A harness that cannot observe a leak fails the
test rather than passing it.
Covered: plain DNS (UDP and TCP), ordinary TCP, DoT on 853, QUIC DoH on UDP/443, ICMP, arbitrary UDP, IPv6 — plus the other direction, that a bypassed UID can still reach the LAN. A firewall that blocked everything would pass the first seven and fail the eighth.
# libpcap is required: the sniffer compiles a BPF filter, and Scapy dlopen()s
# the unversioned libpcap.so that only the -devel/-dev package ships.
sudo apt install nftables iproute2 conntrack libpcap0.8 libpcap-dev # Debian/Ubuntu
sudo dnf install nftables iproute2 conntrack libpcap libpcap-devel # Fedora/RHEL
pip install -e ".[nse]"
make test-nse # runs as root; TTP_REQUIRE_NSE=1 so it cannot skip itself
This runs in CI on every push (the Zero-leak ruleset verification job) and as
a step in scripts/verify.sh before a release.
While Docker integration tests are fast and atomic, they don't capture 100% of the kernel/systemd nuances. For critical changes, it is highly recommended to test in a real QEMU VM:
# Start a specific VM (e.g., arch)
./scripts/vm/start.sh arch
# Sync current code to the VM
./scripts/vm/send.sh
# Snapshot management for easy rollbacks
./scripts/vm/snapshot.sh arch save before-risky-test
If something goes wrong, run the diagnostic command:
sudo ttp diagnose
├── pyproject.toml # Package metadata and dependencies
├── README.md
├── CONTRIBUTING.md # Contribution guidelines
├── SECURITY.md # Security policy
├── scripts/ # Installation, verification, and VM management scripts
├── assets/ # Branding and demo assets
├── packaging/ # Packaging configurations (.deb, .rpm, Arch PKGBUILD)
├── ttp/ # Main Python source package
│ └── resources/ # Internal package resources (SELinux policies, etc.)
├── tests/ # Unit, integration, and leak testing suites
└── docs/ # Technical documentation, threat models, and ADRs
Contributions are welcome, and the areas where help matters most are narrow and specific:
Start with CONTRIBUTING.md, which documents the two rules this codebase is built on: never fix a bug without adding the check that would have caught it, and a test that asserts an absence must first prove it can detect a presence.
| Bugs and feature requests | GitHub Issues |
| Security vulnerabilities | SECURITY.md - please do not open a public issue |
| Version support and EOL | SUPPORT.md |
| Releases and packages | GitHub Releases · PyPI |
This project is maintained in free time. A star helps others find it; sponsorship helps it keep going.
MIT. See LICENSE for more information.