High-Performance Reverse Proxy and WAF for Tor & I2P Hidden Services

MaveWAF is a reverse proxy with integrated WAF designed for Tor and I2P hidden services. Built on Cloudflare's Pingora framework, it provides multi-layer protection:

• Layer 4 (Transport): Circuit-level rate limiting via PROXY protocol
• Layer 7 (Application): Deep HTTP inspection with WAF rules

• Circuit-Based Rate Limiting — Per-circuit throttling via Tor's PROXY protocol
• WAF Engine — SQLi, XSS, Path Traversal, RFI, SSRF detection using libinjection + Aho-Corasick pattern matching
• No-JS CAPTCHA — AI-resistant, works in Tor Browser "Safest" mode
• Adaptive Defense — Auto-escalates based on error rates and circuit floods
• Stateless Sessions — HMAC-SHA256 key derivation + AES-256-GCM encryption, no database required

• Circuit ID Extraction — Via HiddenServiceExportCircuitID haproxy
• Active Defense — Kill circuits and enable PoW via Tor Control Port
• I2P Support — Built-in i2pd integration

• Connection Management — Optimized upstream connection reuse via Pingora's built-in pooling
• Compression — Gzip and Brotli support
• Async I/O — Built on Tokio runtime

graph LR
    A[Tor Browser] --> B[Tor Daemon]
    B -->|PROXY Protocol| C[MaveWAF]
    C -->|Layer 4| D{Rate Limit}
    D -->|Pass| E{WAF Rules}
    E -->|Pass| F[Backend]
    D -->|Block| G[429 Response]
    E -->|Block| H[403 Response]
    C -->|Defense Mode| I[CAPTCHA Challenge]

1. Ingress: Tor forwards traffic with PROXY header containing circuit ID
2. Layer 4: Circuit-based rate limiting before HTTP parsing
3. Layer 7: WAF rules inspect headers, body, and query parameters
4. Decision: Pass, Challenge (CAPTCHA), or Block

# Pull image (Docker Hub)
docker pull mrmave/mavewaf:latest

# Alternative: Pull from Private Forgejo Registry
docker pull git.mrmave.work/maverick/mavewaf:latest

# Configure
cp docs/.env.example .env
# Edit .env with your settings

# Run
docker compose up -d

# Clearnet
git clone https://git.mrmave.work/maverick/mavewaf.git

# Tor / Onion Service
git -c http.proxy=socks5h://127.0.0.1:9050 clone http://mavegitwskioz7tpppmjtj7fn24pwezciii3nvc7kdyltn5iu5uakfqd.onion/mavewaf

cd mavewaf
cargo build --release
./target/release/mavewaf

MaveWAF is configured via environment variables. See the Configuration Reference for all available options.

cp docs/.env.example .env

Your torrc must enable PROXY protocol:

HiddenServiceDir /var/lib/tor/hidden_service/
HiddenServicePort 80 127.0.0.1:8080
HiddenServiceExportCircuitID haproxy

ControlPort 127.0.0.1:9051
HashedControlPassword 16:YOUR_HASHED_PASSWORD

Generate Tor control password:

tor --hash-password "your_password"

Standard alphanumeric challenge with slight rotation. Good for most use cases.

High-entropy, AI-resistant challenge with:

• Aggressive visual noise
• Decoy characters
• Geometric patterns

Inspired by "Endgame DDoS" protection systems.

services:
  mavewaf:
    image: mrmave/mavewaf:latest
    restart: unless-stopped
    env_file:
      - .env
    volumes:
      - ./tor_keys:/var/lib/tor/hidden_service/
      - ./torrc:/etc/tor/torrc:ro
      - ./i2p_keys:/var/lib/i2pd/

RUST_LOG=warn           # Production
RUST_LOG=debug          # Development
LOG_FORMAT=json         # Structured output

All security events include: circuit_id, http_method, http_path, action, rule.

cosign verify --key cosign.pub git.mrmave.work/maverick/mavewaf:latest
cosign verify --key cosign.pub mrmave/mavewaf:latest

wget https://git.mrmave.work/maverick/mavewaf/releases/download/vX.X.X/mavewaf-vX.X.X-linux-amd64
wget https://git.mrmave.work/maverick/mavewaf/releases/download/vX.X.X/mavewaf-vX.X.X-linux-amd64.sha256
wget https://git.mrmave.work/maverick/mavewaf/releases/download/vX.X.X/mavewaf-vX.X.X-linux-amd64.asc

# Verify using WKD (Web Key Directory)
gpg --locate-keys [email protected]

# Or import manually
wget https://git.mrmave.work/maverick/mavewaf/raw/branch/master/maverick.asc
gpg --import maverick.asc

sha256sum -c mavewaf-vX.X.X-linux-amd64.sha256
gpg --verify mavewaf-vX.X.X-linux-amd64.asc mavewaf-vX.X.X-linux-amd64

• Onion Balancer: Currently not supported. This is planned for future development to allow scaling across multiple backend instances.
• File Uploads: Fully supported via Content-Length. Large files are buffered in memory for security inspection.

AGPL-3.0