Skip to content

Security: rlnunez/OpenOptOut

SECURITY.md

Security policy

OpenOptOut handles sensitive patron and family data. Security reports are welcome.

Reporting a vulnerability

Please do not open a public issue for a security vulnerability.

Report privately via Report a vulnerability (Security tab → Report a vulnerability). This project implements RFC 9116 (/.well-known/security.txt).

Include:

  • Vulnerability description and potential security impact
  • Reproduction steps (requests, scripts, or screenshots)
  • Affected version or commit hash (available via /api/health)

You should receive an initial response within 7 days. Valid reports will be credited in published advisories, the Credits, and the Changelog unless anonymity is requested.

Scope

  • In scope: Backend API, web UI, plugin sandbox (bubblewrap, seccomp, gRPC broker), data encryption at rest, authentication mechanisms (local, OIDC, SAML, LDAP, SIP2), distributed workers, installers, and official Docker images.
  • Out of scope: Third-party data broker sites, and issues requiring prior root or host compromise (unless enabling container escapes or cross-tenant privilege escalation).

Security Researcher Safe Harbor

In accordance with the CISA "Secure by Design" Product Principles:

  • Good-Faith Research: We consider security research conducted under this policy to be authorized. If you make a good-faith effort to comply with this policy during your security research, we will not initiate or pursue legal action against you (including claims under the Computer Fraud and Abuse Act (CFAA) or DMCA anti-circumvention provisions).
  • Guidelines for Researchers:
    • Test only against your own self-hosted instances or local test environments. Never attempt to access or modify another family's or patron's personal data.
    • Do not execute denial-of-service (DoS) attacks, automated high-rate fuzzing against production infrastructure, or destructive exploits.
    • Never test against third-party data broker sites using real personal details. Use synthetic details (e.g. Jordan Example, @example.invalid) and safe form rehearsal modes.
    • Give the project reasonable time to remediate reported vulnerabilities before public disclosure.

Core Data Protection & Privacy Controls

  • No Personal Data Outside Encrypted Storage: Patron and family member identifiers (names, addresses, phones, emails) are encrypted at rest using AES-256-GCM (NIST SP 800-38D) with HKDF domain separation.
  • Never Store SSNs or Government IDs: Hard architectural boundary. SSNs, government IDs, and third-party passwords are never collected, filled, or stored.
  • Children's Data Protection (COPPA): Dependent profiles are managed exclusively through authenticated parent/guardian accounts. Aux data (confirmation links, email logs) for minors is automatically crypto-shredded after the retention window (coppa_children_retention_days).
  • Zero Third-Party Telemetry: Zero analytics SDKs, tracking pixels, or outbound telemetry. Outbound network requests are limited to admin-configured integrations and broker automation.
  • Memory Hygiene & Crash Protection: Core dumps are suppressed at the process (RLIMIT_CORE = 0), container (ulimits: core: 0), and service levels to prevent cleartext memory dumps to disk. Ephemeral decrypted buffers are zeroized upon completion.
  • Plugin Sandboxing: Broker plugins execute within OS-level namespaces (bubblewrap), seccomp syscall filters, and declared network boundaries.

Data Classification

Data Category Classification Encryption & Protection Retention & Access
Family Member PII Restricted (COPPA) AES-256-GCM at rest; access logged; scrubbed from all logs Self-service deletion; minors' aux data auto-purged
SSNs & Government IDs Prohibited Never collected, filled, or stored N/A
Credentials & Secrets Restricted AES-256-GCM / Fernet; masked in UI; never logged Kept until replaced/removed; admin only
Session Tokens (JWT) Restricted Signed (HMAC-SHA256); ephemeral browser storage Revocable per-session or account-wide; expires per policy
Security Audit Logs Internal Structured text (0600 permissions); PII scrubbed Configurable retention (default 30+ days); admin only
Interaction Artifacts Restricted PII encrypted; CAPTCHAs blacked out before transmission Purged per retention schedule
Uploaded Branding Low Validated, stripped of metadata, re-encoded from pixels Replaced on upload; public login asset

Web Frontend & Browser Security

  • Token Authentication: Authentication uses Bearer tokens passed in the Authorization header rather than ambient cookies, eliminating cross-site request forgery (CSRF) exposure.
  • Security Headers & CSP: Enforces nosniff, strict-origin-when-cross-origin, frame restrictions, and an explicit Content-Security-Policy (object-src 'none', base-uri 'none', frame-ancestors 'none') with reporting to /api/csp-report.
  • Self-Hosted Assets: All JavaScript, stylesheets, and fonts are bundled locally with no external CDN runtime dependencies.
  • Input Neutralization: External links and broker text are validated and sanitized (isSafeHref, control character filtering) to prevent injection and malicious scheme execution.

Authentication & Session Controls

  • Progressive Throttling: Exponential backoff and independent account/IP locking on repeated failed attempts without account enumeration.
  • Multi-Factor Authentication: TOTP (RFC 6238) and FIDO2/WebAuthn with counter-replay protection and password re-verification.
  • Session Tracking & Revocation: Each sign-in issues a discrete session ID (UserSession). Users can view and revoke individual sessions or sign out everywhere; admins can force-revoke sessions globally.
  • Audit Logging: Structured audit lines (auth_event, authz_denied, lockout_triggered) record security events in UTC with numeric IDs and scrubbed PII.

Cryptographic Inventory & Key Management

All cryptographic keys are operator-supplied via environment variables or owner-only files (0600), or derived using RFC 5869 HKDF for purpose-specific domain separation.

Key / Material Purpose Algorithm / Standard Rotation Guidance
SECRET_KEY Root secret, JWT signing, HKDF root HMAC-SHA256 (128+ chars) Annually, or immediately on suspected compromise
FIELD_ENCRYPTION_KEY Patron PII encryption at rest AES-256-GCM (NIST SP 800-38D) On compromise (with data re-encryption)
DB_ENCRYPTION_KEY Database file encryption SQLCipher AES-256 On compromise
Password Pepper Pre-hash passwords before bcrypt PBKDF2-HMAC-SHA256 (210k rounds) On compromise (independent of SECRET_KEY)
SEARXNG_SECRET Search instance session signing HMAC-SHA256 Annually
Queue HMAC Key Distributed worker envelope integrity HMAC-SHA256 (HKDF-derived) Derived from SECRET_KEY
Backup Key Backup archive confidentiality AES-256-GCM (scrypt/PBKDF2) Per organizational backup policy
DB / Redis Passwords Service-to-service authentication High-entropy string Annually or on staff turnover

External Services & File Uploads

  • External Integrations: Outbound connections (SMTP/IMAP, SearXNG, SSO IdPs, webhooks, brokers) are strictly administrator-configured, enforce short bounded timeouts (10–25s), and do not silently cascade retries.
  • File Upload Boundaries: All upload endpoints are restricted to authenticated administrators and enforce size caps:
    • Branding logos: 2 MB limit; magic-byte verified; raster images re-encoded from raw pixels; SVGs sanitized.
    • Broker plugins: 50 MB zip limit; manifest verified; inspected for path traversal and disallowed syscalls before installation.
    • CSV/JSON/Registry imports: 10–30 MB limits; strict schema validation; uncompressed size and entry limits enforced on zip/xlsx archives.

Compliance & Security Standards

OpenOptOut is engineered to follow recognized industry standards for application security, cryptography, and privacy:

  • NIST SP 800-38D: Authenticated Encryption with Galois/Counter Mode (AES-256-GCM).
  • NIST SP 800-132 & SP 800-63B: Password-Based Key Derivation (PBKDF2), MFA standards, and session controls.
  • FIDO Alliance FIDO2 / W3C WebAuthn Level 3: Hardware security keys, User Verification (UV) enforcement, Backup Eligibility (BE) flag classification, and Ed25519 (EdDSA) curves.
  • CISA / NSA / FBI Secure by Design: Principles and Approaches for Secure by Design Software, vulnerability class elimination, researcher safe harbor, and RFC 9116 (see docs/architecture/cisa-secure-by-design.md).
  • NIST Privacy Framework Version 1.0: Enterprise Privacy Risk Management and Cybersecurity Framework (CSF) Privacy Profile (see docs/architecture/nist-privacy-profile.md).
  • ISO/IEC 29100: Privacy Architecture Framework, role boundaries, and 11 privacy principles (see docs/architecture/privacy-framework-29100.md).
  • ISO/IEC 27701 & COPPA: Privacy Information Management, parental consent boundaries, and automated crypto-shredding.
  • OWASP ASVS 5.0.0 Level 2: Application Security Verification Standard baseline for web applications and APIs.
  • OWASP Top 10 for LLMs: Large Language Model application security and prompt injection encapsulation boundaries.
  • WCAG 2.2 Level AAA: Universal web accessibility and contrast (see ACCESSIBILITY.md).

Automated Scanning

Every change is verified by automated static and dynamic security scanners. See docs/dev/security-scanning.md.

There aren't any published security advisories