Skip to content

Repository navigation

Security: savfox-ai/savfox

Security

SECURITY.md

Security Policy

Supported Versions

Version Supported
0.3.x ✅
< 0.3 ❌

Reporting a Vulnerability

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

  1. Do NOT open a public GitHub issue
  2. Email: chris@sonc.ai
  3. Include:
    • Description of the vulnerability
    • Steps to reproduce
    • Potential impact
    • Suggested fix (if any)

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

Security Model

Gateway Server

  • Token-based authentication with scoped permissions (Operator, Viewer, Chat)
  • Rate limiting on authentication attempts (token-bucket algorithm)
  • Non-root execution in Docker containers
  • TLS support for encrypted connections

Agent Execution

  • Sandboxed execution via platform-native mechanisms:
    • Linux: Landlock + seccomp
    • macOS: Seatbelt (sandbox-exec)
    • Windows: Restricted token
  • Execution approval workflow for sensitive commands
  • Tool policy enforcement (allow/deny lists)

Data Protection

  • Credential storage via OS keyring (keyring-store crate)
  • No plaintext secrets in configuration files
  • Session data stored locally with file permissions

Threat Model

In Scope

  • Remote code execution via gateway WebSocket
  • Authentication bypass
  • Privilege escalation in sandbox
  • Information disclosure via API endpoints
  • Denial of service via resource exhaustion

Out of Scope

  • Physical access to the host machine
  • Social engineering of users
  • Vulnerabilities in upstream LLM providers
  • Vulnerabilities in third-party chat platforms (Discord, Slack, etc.)

There aren't any published security advisories