If you discover a security vulnerability, please report it responsibly:

1. Do NOT open a public issue
2. Email details to the maintainers via GitHub Security Advisories
3. Include steps to reproduce, impact assessment, and any suggested fixes

We will acknowledge receipt within 48 hours and aim to release a fix within 7 days for critical issues.

• Azure Static Web Apps with global CDN and DDoS protection
• Cloudflare (Project Galileo) provides additional CDN, WAF, and DDoS mitigation
• HTTPS/TLS 1.3 enforced with HSTS preload (1 year)
• Tor hidden service available for censorship-resistant access

Strict CSP headers limit resource loading to explicitly allowed origins:

• default-src 'self'
• connect-src allowlist: Azure services, Typesense search, Carl AI endpoints, Plausible analytics
• frame-ancestors 'none' (no embedding)
• upgrade-insecure-requests enforced

• X-Content-Type-Options: nosniff
• X-Frame-Options: DENY
• Referrer-Policy: strict-origin-when-cross-origin
• Cross-Origin-Opener-Policy: same-origin
• Cross-Origin-Embedder-Policy: credentialless
• Permissions-Policy restricts camera, microphone, geolocation, payment APIs

• CodeQL analysis on every push and PR, plus weekly scheduled scans
• OSV-Scanner for dependency vulnerability detection on PRs and weekly
• Dependabot automated dependency updates with auto-merge for patches
• Pre-commit hooks via Husky for local linting

• No user accounts, no cookies, no tracking pixels
• No personal data collected or stored
• Self-hosted Plausible Analytics — aggregate metrics only, no IPs or identifiers
• GPS coordinates never leave the device — all distance calculations are client-side
• Mobile crash reporting (Sentry) is opt-in only

• Self-hosted on dedicated hardware — Ollama + vLLM run on a Mac Mini, not third-party cloud AI
• Queries are not stored or used for training
• No PII is sent to AI endpoints — client-side sanitization before any LLM call
• AI endpoints accessible via Tor for maximum privacy

• Function-level authorization keys for sensitive endpoints
• Azure managed identities where possible
• No secrets in code — all credentials via GitHub Secrets or Azure Key Vault

The following are in scope for security reports:

• baynavigator.org (web application)
• Azure Functions endpoints
• Carl AI endpoints (ai.baytides.org, ollama.baytides.org)
• Mobile applications (iOS, Android)
• GitHub Actions workflows
• Infrastructure-as-code (Bicep templates)

• Third-party services we link to (program websites)
• Plausible Analytics instance
• Denial of service attacks