Skip to content

SECURITY_ASSESSMENT.md

Latest commit

 

History

History
108 lines (75 loc) · 3.27 KB

SECURITY_ASSESSMENT.md

File metadata and controls

108 lines (75 loc) · 3.27 KB

Ironclaw Security Assessment

Date: March 21, 2026 | Posture: Good (7.5/10) | Critical Issues: 0

Executive Summary

Ironclaw has strong security foundations: AES-256-GCM encryption, Docker sandboxing, zero-exposure credential model, and defense-in-depth architecture. Priority improvements: secure .env permissions and migrate tokens to secrets manager.

Current Security Posture

✅ Strengths

Area Implementation
Secrets Management AES-256-GCM encryption, OS keychain integration, per-secret key derivation (HKDF-SHA256), leak detection (15+ patterns)
Docker Sandbox Non-root (UID 1000), cap_drop: ALL, read-only root FS, tmpfs mounts, network allowlist, resource limits
API Security Bearer token auth (constant-time), rate limiting (30/60s chat, 60/min webhook), HMAC-SHA256 signatures, CORS restrictions
Safety Layer Input sanitization, injection detection, policy enforcement, environment scrubbing

⚠️ Priority Improvements

Priority Issue Mitigation
HIGH .env file plaintext storage Migrate to secrets manager, chmod 600
MEDIUM SANDBOX_POLICY=workspace_write Appropriate for personal use; never use full_access
MEDIUM No security audit logging Add RUST_LOG=ironclaw::secrets=debug
LOW Database SSL prefer mode Use require for remote PostgreSQL

Key Findings

Secrets Flow

~/.ironclaw/.env → Ironclaw reads → AES-256-GCM encrypt → PostgreSQL (encrypted) → HTTP proxy injects → External APIs

Risk Assessment

Threat Likelihood Impact Status
.env file theft Medium High 🟡 Mitigate with chmod 600 + secrets manager
Prompt injection High Medium ✅ Safety layer + sandbox
Sandbox escape Low Critical ✅ Docker hardening
Credential leak Low High ✅ Zero-exposure model
Network MITM Low High ✅ HTTPS, localhost DB

Overall Risk: 🟢 LOW (appropriate for personal use)

Immediate Actions (< 30 min)

# 1. Secure .env permissions
chmod 600 ~/.ironclaw/.env
chmod 700 ~/.ironclaw

# 2. Verify sandbox policy
grep SANDBOX_POLICY ~/.ironclaw/.env  # Should be: workspace_write

# 3. Check master key status
./target/release/ironclaw doctor | grep -A2 "Master Key"

# 4. Migrate tokens to secrets manager
./target/release/ironclaw secrets set telegram_bot_token "YOUR_TOKEN"
./target/release/ironclaw secrets list

Verification Commands

# File permissions
ls -la ~/.ironclaw/.env  # Should show: -rw-------

# Sandbox status
./target/release/ironclaw sandbox status

# Secret access test
./target/release/ironclaw sandbox test --network api.telegram.org

# Recent security events
tail -100 /tmp/ironclaw.log | grep -iE "error|warn|secret"

Architecture Overview

User Input → Auth (Bearer Token) → Rate Limiting → Safety Layer (Injection/Leak Detection)
    ↓
Docker Sandbox (Non-root, Cap Drop, Network Proxy)
    ↓
Secrets (Encrypted at Rest, Injected at HTTP Proxy)
    ↓
External APIs (HTTPS Only)

Next Review: June 21, 2026 (quarterly) | Full docs: See original assessment in git history