| Version | Supported |
|---|---|
| 0.7.x | ✅ Yes |
| below 0.7 | ❌ No |
Do NOT open a public GitHub issue for security vulnerabilities.
Email: support@astralabs-ai.net
Please include:
- Description of the vulnerability
- Steps to reproduce
- Potential impact assessment
- Suggested remediation (if known)
We respond within 48 hours and patch critical issues within 7 days.
| Control | Implementation |
|---|---|
| Authentication | JWT HS256, 1h default expiry; rotating refresh tokens |
| Passwords | bcrypt 12 rounds |
| Rate limiting | 10/min login, 5/min register |
| Security headers | CSP, X-Frame-Options, nosniff; HSTS on production HTTPS |
| CORS | Allowlist only |
| SQL injection | Parameterised queries only |
| Evidence integrity | SHA-256 hash chain |
| Tenant isolation | org_id scoping on all queries |
| LLM calls | core/llm multi-provider router (Anthropic, OpenAI, stub); CircuitBreaker on ingest; no document body in audit payloads |
| Supply chain (CI) | pip-audit and npm audit (high severity threshold for npm) on push/PR to main |
| Variable | Purpose |
|---|---|
CORTEX_LLM_PROVIDERS |
Comma-separated chain, e.g. anthropic,openai,stub (default) |
ANTHROPIC_API_KEY |
Anthropic Messages API |
ANTHROPIC_MODEL |
Model id (default claude-sonnet-4-20250514) |
OPENAI_API_KEY |
Optional OpenAI fallback |
OPENAI_MODEL |
OpenAI model (default gpt-4o-mini) |
Before release, run the living checklist in docs/RELEASE_QA.md.
Historical snapshots (not current posture) live under docs/archive/:
| Report | Purpose |
|---|---|
| docs/archive/QA-REPORT.md | Dated QA / smoke log |
| docs/archive/SAST-REPORT.md | Dated SAST / dependency audit |
| docs/archive/SECURITY_REPORT.md | Dated security posture note |
- JWT_SECRET default — not production-safe. Always set
JWT_SECRETenvironment variable. - CORS — set
FRONTEND_URLenv var to your production domain. - HTTPS — terminate TLS at the production proxy/CDN and set
CORTEX_ENVIRONMENT=productionplusCORTEX_HTTPS_ENABLED=1. - Frontend headers — the API middleware cannot secure a separately served SPA.
Configure CSP and HSTS on the HTTPS proxy/CDN serving
frontend/, validate the policy in report-only mode, then enforce it.
Astra GRC governs its own AI under EU AI Act Art.14. Every automated compliance decision with confidence below 0.75 routes through the Human Review Queue. The Evidence Vault (SHA-256 hash chain) provides the audit trail required under NIS2 Art.20.