If you discover a security vulnerability in CryptoTrace, do not open a public GitHub issue. Please report it privately.

Send an encrypted email to security@cryptotrace.dev using our PGP key:

Fingerprint: 3A4B 5C6D 7E8F 9A0B 1C2D 3E4F 5A6B 7C8D 9E0F 1A2B

If you do not have PGP set up, you may report via the private vulnerability reporting mechanism on GitHub (if enabled for this repository).

• 48 hours: Initial acknowledgement of receipt
• 7 days: Triage and severity assessment with expected fix timeline
• 90 days: Standard embargo period from disclosure to public fix
  • Extensions may be granted for complex vulnerabilities

• Affected version(s) and platform(s)
• Description of the vulnerability and impact
• Steps to reproduce (PoC preferred but not required)
• Any suggested mitigation or fix

The following are in scope for security reports:

• CryptoTrace engine (Rust binaries)
• Signature database loading and parsing
• Input sanitization and sandboxing
• REST API authentication and rate limiting
• AI provider integration and prompt handling
• Update mechanism and signature verification

The following are out of scope:

• Third-party AI provider infrastructure (OpenAI, Anthropic, Ollama)
• Operating system vulnerabilities
• Social engineering of project maintainers

Risky parsing operations (decompression, binary parsing) run in isolated subprocesses via:

• Windows: Job Objects with process group kill on timeout
• Linux: Seccomp-bpf with restricted syscall whitelist
• macOS: sandbox-exec with minimal entitlements

If a worker process crashes or is compromised, the main analysis pipeline remains unaffected.

All cloud/AI features are opt-in. The engine operates fully air-gapped with no network calls unless explicitly configured. This is verified by:

• No HTTP client initialized at startup
• No DNS resolution attempted during normal operation
• All network features gated behind explicit configuration flags
• Network calls logged at info level for auditability

• Structured output only (no freeform text generation)
• Per-field JSON validation prevents hallucination
• All AI output flagged as non-authoritative
• Prompt injection defense: AI never receives raw user input, only structured DetectionResult JSON
• Temperature set to 0.1 (configurable) to reduce injection success rate

• Signature database updates verified via Ed25519 signatures (primary) or GPG (fallback)
• Dual-repo architecture: engine and signature sources are independently maintained
• Provenance log tracks every update with timestamp and fingerprint
• Rollback capability: cryptotrace update --rollback
• Air-gap import: cryptotrace update --from-file <path> --verify <sig>

• API keys stored in config file or environment variables, never in code
• cryptotrace config show redacts all secrets
• .gitignore covers all secret file patterns
• detect-secrets pre-commit hook recommended for contributors