Skip to content

Security: hyperpolymath/llm-unify

SECURITY.md

Security Policy

Reporting Security Vulnerabilities

We take security seriously. If you discover a security vulnerability in LLM Unify, please report it responsibly.

Contact

Email: security@llm-unify.dev PGP Key: See .well-known/security.txt

Response Timeline

  • Critical vulnerabilities (RCE, privilege escalation, data exposure): 7 days
  • High severity (authentication bypass, injection): 14 days
  • Medium severity (DoS, information disclosure): 30 days
  • Low severity (configuration issues, minor leaks): 60 days

What to Include

  1. Description of the vulnerability
  2. Steps to reproduce the issue
  3. Proof of concept (if applicable)
  4. Impact assessment (what can be compromised)
  5. Suggested fix (if you have one)

Security Considerations

Local-First Architecture

LLM Unify is designed with a local-first architecture:

  • No network calls - All data stays on your machine
  • SQLite storage - Encrypted at rest with user's file system permissions
  • No telemetry - Zero data collection or phone-home behavior

Known Security Measures

  1. SQL Injection Prevention

    • All database queries use parameterized statements (SQLx)
    • No raw SQL string concatenation
    • Type-safe query builders
  2. Input Validation

    • File path sanitization for import/export operations
    • JSON parsing with strict schema validation
    • Message content size limits (configurable)
  3. Memory Safety

    • Zero unsafe blocks - Rust compile-time guarantees
    • No manual memory management
    • Thread-safe concurrency (SQLite connection pooling)
  4. Dependency Management

    • Regular cargo audit scans in CI/CD
    • Minimal dependency footprint
    • Vetted crates only (serde, sqlx, ratatui, clap)

Known Limitations

  1. Database Encryption

    • SQLite database is not encrypted by default in v0.1
    • Relies on file system permissions
    • Plan to add encryption in v0.2 (see roadmap)
  2. Export Format

    • Exported JSON files contain plaintext conversations
    • Users should secure export files appropriately
  3. Parser Trust

    • Import parsers trust input data structure
    • Malformed exports may cause parse errors (non-exploitable)

Security Audit History

Date Auditor Scope Findings Status
2025-11-28 Internal Initial release None Clean

External audit: Planned for v0.2.0 release (Q1 2025)

Security Features Roadmap

v0.2.0 (Planned Q1 2025)

  • SQLite database encryption (SQLCipher integration)
  • Export file encryption (age/GPG options)
  • Secure key management
  • Formal threat modeling

v0.3.0 (Planned Q2 2025)

  • External security audit
  • Penetration testing results
  • Security.txt automation

Supported Versions

Version Supported
0.1.x
< 0.1

Security Best Practices for Users

  1. File Permissions: Protect your database file (llm-unify.db)

    chmod 600 llm-unify.db
  2. Backup Security: Encrypt backups before cloud storage

    llm-unify backup backup.db
    gpg -c backup.db  # Encrypt with passphrase
  3. Import Safety: Only import exports from trusted sources

  4. Update Regularly: Keep LLM Unify updated for security patches

Compliance

This security policy follows:

  • RFC 9116 - security.txt standard
  • OWASP Top 10 - Web application security risks
  • Rhodium Standard Repository - Security documentation requirements

There aren’t any published security advisories