Skip to content

Security: BrianNguyen29/Ferrum-Gate

Security

SECURITY.md

Security Policy

FerrumGate exists to improve execution governance for agentic systems. We take security seriously and welcome responsible disclosure.

Supported Versions

Version Notes
v0.1.0 Current development version; single-node SQLite focus
< v0.1.0 Unsupported

Reporting a Vulnerability

Please use private disclosure first. Do not open a public issue for unpatched vulnerabilities.

To report a security issue, email the maintainers directly or use GitHub Private Vulnerability Reporting if enabled. Include:

  • A clear description of the vulnerability
  • Steps to reproduce (or a plausible attack scenario)
  • Affected versions or commit ranges
  • Any suggested mitigations

Response Expectations

  • Acknowledgment: Within 5 business days
  • Initial assessment: Within 10 business days
  • Fix timeline: Target 30 days for confirmed vulnerabilities; complex issues will receive a public timeline update
  • Disclosure: Coordinated disclosure preferred; we will work with reporters to agree on a reasonable publication date

Scope of Security Review

The following areas are in scope for vulnerability reports:

  • Capability misuse or escalation
  • Policy bypass in the PDP/gateway
  • Taint / provenance gaps that could lead to unsanctioned execution
  • Rollback or compensation failures that break reversibility guarantees
  • Integrity mismatches in lineage or ledger records
  • Output sanitization leaks
  • Authentication or authorization weaknesses in the bearer-token gate
  • Rate-limit bypass

Out of Scope

  • Denial of service via resource exhaustion without a reproducible exploit
  • Issues in dependencies unless they directly affect FerrumGate's invariants
  • Social engineering or physical attacks
  • Vulnerabilities in unsupported versions

No Bug Bounty

FerrumGate does not operate a bug bounty program. We appreciate responsible disclosure and will acknowledge reporters in release notes unless they prefer anonymity.

Security Model

  • Intent-scoped execution with capability bounds
  • Approval-aware workflow with lineage tracking
  • Rollback-classified operations (compensate is the v1 recovery endpoint)
  • Bearer-token auth in bearer-auth mode; dev config remains auth-disabled for local development
  • Rate limiting integrated with the gateway

There aren't any published security advisories