Thank you for taking the time to help make PePFibPred more secure.

Do not open a public GitHub issue for a security vulnerability.

Instead, report privately to:

• Primary contact: Said Azaizah — said.azaizah@cssb-hamburg.de (DESY-CSSB)
• Secondary contact: Aleksandr Golubev — aleksandr.golubev@cssb-hamburg.de (DESY-CSSB)
• GitHub Security Advisories: https://github.com/az-said/peptide_prediction/security/advisories/new

Please include:

1. A description of the vulnerability
2. Steps to reproduce (a minimal proof-of-concept is ideal)
3. The affected version(s)
4. Any suggested mitigations you already know
5. Whether you plan to disclose publicly and on what timeline

We commit to:

• Acknowledging your report within 3 business days
• A working fix in the codebase for Critical issues within 48 hours of confirmation
• A working fix for High issues within 7 days
• A working fix for Medium / Low issues within 30 days or the next planned release, whichever is sooner
• A public advisory + credit (unless you prefer anonymity) once the fix ships

In scope

• The pvl backend (backend/) and its FastAPI HTTP surface
• The pvl-cli Python package published to PyPI
• The pvl-mcp Model Context Protocol server published to PyPI
• The container images published to ghcr.io/az-said/peptide_prediction
• The reference deployment at the current production hostname

Out of scope

• Vulnerabilities in third-party dependencies for which no PePFibPred configuration is exploitable (report to the upstream)
• Vulnerabilities in TANGO or S4PRED themselves that are not exploitable through PePFibPred's exposed surfaces
• Denial-of-service via unbounded input at the /api/predict/batch route beyond the 30-req/min/IP rate limit (this is a known-and-documented rate-limiting boundary, not a vulnerability)

PePFibPred is built with the following supply-chain controls:

• Dependabot — weekly grouped updates across pip, npm, GitHub Actions, and Docker (.github/dependabot.yml)
• CodeQL — security-extended queries run on every pull request and on a weekly cron (.github/workflows/codeql.yml)
• Trusted Publishing — PyPI OIDC / OpenID-Connect for pvl-cli and pvl-mcp (.github/workflows/publish-pypi.yml) — no long-lived credentials leave the GitHub Actions environment
• Build provenance attestations — SLSA Build Level 2 for both container images (actions/attest-build-provenance@v4 in .github/workflows/docker-publish.yml)
• Secret scanning + push protection — GitHub-native, enabled at the org level
• Vulnerability alerts — GitHub Advisory Database → Dependabot alerts, enabled

The pvl cache uses SHA-256 (truncated to 128 bits) as a sequence-normalisation key for deterministic per-predictor caching. It is not intended as a cryptographic authenticity check.

The Sentry release integration signs source maps at upload time. TLS is terminated by Caddy 2 with automatic Let's Encrypt certificate rotation.

Security researchers who responsibly disclose vulnerabilities will be credited by name in the security advisory unless anonymity is requested.