Skip to content

Security: eggstack/eggress

Security

SECURITY.md

Security

Supported Versions

Version Supported
Current (main) Yes
< Current No

Eggress follows a rolling release model. Only the latest release receives security fixes.

Reporting a Vulnerability

Report security issues via GitHub Issues using a private disclosure if possible (GitHub supports confidential issues for repo collaborators).

Expected response timeline:

  • Acknowledgment: within 72 hours
  • Triage and severity assessment: within 7 days
  • Fix or mitigation: depends on severity, but critical issues are prioritized for the next release

Do not disclose vulnerabilities publicly until a fix is available.

Security Features

  • unsafe_code = "deny" in all workspace crates — no memory safety issues possible through Rust code
  • No OpenSSL — uses rustls with ring crypto provider, eliminating C FFI attack surface
  • Credential redaction — RedactedUri replaces credentials with ****:****@ in all logs, metrics, admin output, and error messages
  • UDP amplification prevention — validate_target() rejects multicast, broadcast, and unspecified addresses
  • UDP client pinning — prevents address spoofing for UDP association ownership
  • HTTP header injection prevention — control character validation in proxy credentials
  • Config validation — structural errors, duplicate names, invalid references, and dangerous binds are rejected at load time
  • Admin loopback default — admin server binds to 127.0.0.1 unless explicitly configured otherwise
  • Non-loopback bind warnings — config validation emits warnings for unauthenticated network-facing listeners
  • Atomic config reload — lock-free reads via ArcSwap; hot-reloadable fields swapped atomically
  • Bounded parsing — HTTP headers, SOCKS fields, and credential inputs have explicit size limits

Documentation

Detailed security documentation lives in docs/security/:

Scope

Eggress is a single-operator, controlled-network proxy toolkit. The threat model assumes a trusted operator managing configuration on a trusted host. Untrusted inputs are network clients, upstream proxy responses, and parsed config file contents. See docs/security/THREAT_MODEL.md for the full threat model.

There aren't any published security advisories