Skip to content

Security: g8e-ai/g8e

.github/SECURITY.md

Security Policy

Project: g8e — Byzantine Fault Tolerant Governance Platform
Maintained by: Lateralus Labs


Supported Versions

Version Supported
main (latest)
Older releases ❌ — upgrade to latest

Reporting a Vulnerability

Do not open a public GitHub issue for security vulnerabilities.

Report privately to: security@lateraluslabs.com

Include as much of the following as you can:

  • Description of the vulnerability and its potential impact
  • Affected component(s): g8e Gateway (PDP), g8e Operator (PEP), g8e Protocol
  • Steps to reproduce or a minimal proof-of-concept
  • Your assessment of severity (Critical / High / Medium / Low)
  • Whether you believe the issue is currently being exploited

We will acknowledge receipt within 48 hours and provide an initial assessment within 5 business days.


Disclosure Policy

g8e follows coordinated disclosure:

  1. You report privately to us.
  2. We confirm, assess, and develop a fix.
  3. We release the fix and credit you (unless you prefer anonymity).
  4. You may publish after the fix is released, or after 90 days from initial report — whichever comes first.

We will not pursue legal action against researchers acting in good faith under this policy.


Scope

In Scope

  • g8e Gateway (PDP)GovernanceEnvelope parsing, deterministic transaction hash binding, L1-L4 verification logic (Doctrine, Consensus, Notary, Warden), WebAuthn/Passkey L3 brokerage
  • g8e Operator (PEP) — execution boundary (L5 Actuator), signed ActionReceipt issuance, mTLS tunnel, SQLAuditStore, Sovereign Execution Boundary (scrubbing, rehydration)
  • g8e Protocol — Protobuf schemas (common.proto, operator.proto), canonical JSON (protojson) serialization, envelope integrity
  • Authentication — WebAuthn/FIDO2 L3 Notary flow, Ed25519 signature verification (L2 Consensus, L4 Warden, L5 Actuator), replay protection (Nonce)
  • CLI and bootstrapg8e auth login, mTLS credential and PKI handling

Out of Scope

  • Third-party model providers (Anthropic, OpenAI, etc.)
  • Vulnerabilities in dependencies that have already been publicly disclosed and are pending upstream fix
  • Social engineering or phishing attacks against Lateralus Labs employees
  • Denial-of-service attacks without demonstrated security impact beyond availability

Security Architecture Notes

The following are structural properties of g8e, provided to help researchers understand the intended security model:

  • Fail-closed by design. Any verification failure at the L1-L4 layers (Doctrine, Consensus, Notary, Warden) drops the payload and writes an audit record. There is no fallback execution path.
  • Sovereign Execution Boundary. The g8e Operator (PEP) acts as the sovereign boundary. It refuses to mutate host reality unless a transaction carries a valid L2 signature (Tribunal consensus) and, where required, an L3 proof (Human-in-the-loop).
  • No ambient execution authority. No component holds standing permission to mutate state. Authority is granted strictly per-transaction via the GovernanceEnvelope, verified independently at the PEP.
  • Local audit sovereignty. Raw forensic material is stored locally in the SQLAuditStore. The Sovereign Execution Boundary scrubs all outbound data before delivery to remote clients or AI systems.
  • Mandatory encryption at rest. All storage services require an unlocked vault for initialization. Sensitive data (command stdout/stderr, file diffs, content) is encrypted at rest using AES-256-GCM with per-operation nonces. Encryption operations fail-closed if the vault is locked.
  • mTLS everywhere. All platform communication (Operator-to-Gateway) requires mutual TLS. Unauthenticated or unverified connections are rejected.
  • State & Replay Protection. Transactions are bound to a state_merkle_root, protected by a unique nonce, and carry a temporal expires_at deadline.

If your finding demonstrates a bypass of any of these properties, treat it as Critical.


CVE and Dependency Scanning

g8e runs automated dependency scanning on every build. If you identify a dependency vulnerability not yet captured by our tooling, please report it via the channel above.


Hall of Fame

We gratefully acknowledge security researchers who responsibly disclose vulnerabilities. With your permission, your name or handle will be listed here.

No entries yet — be the first.


Contact

Security: security@lateraluslabs.com
General: hello@lateraluslabs.com
Website: https://lateraluslabs.com

There aren't any published security advisories