Skip to content

Security: FeirAI/averin

SECURITY.md

Security policy

averin is alpha software. Its assurance claims are bounded by docs/coverage-limits.md and the threat model in docs/dev/SECURITY.md. Reading those first will help you tell an intended limit from a defect, but if you are unsure, report it.

Reporting a vulnerability

Report privately to security@feir.ai. Do not open a public issue for anything exploitable.

Include what you can of: the affected component (core/, server/, verifier/, web/, sdk/), the commit or tag, steps to reproduce, and the impact you believe it has on the record-integrity, event-observation, or action-accountability claims.

You will get an acknowledgement within 7 days. There is no bug bounty.

What we especially want to hear about

  • A documented limit that does not hold the way the documents say it does.
  • Any way to bypass a guard, including the production-secrets gate that is enabled by setting AVERIN_REQUIRE_PROD_SECRETS to 1 or true, or any way a globally-known development value is accepted while that gate is on.
  • Unexpected impact from a behaviour that is documented as a non-goal.

What is out of scope

Reports that solely restate a limit already documented in docs/coverage-limits.md or under "What averin deliberately does NOT do" in docs/dev/SECURITY.md, with no new impact. The development signing seed in spec/golden-vectors/sign-vectors.json is public by construction; a report that it is usable as a key is expected, a report that it is accepted while the production-secrets gate is on is not.

Do not run denial-of-service or load tests against instances you do not operate. Availability weaknesses found without such testing are welcome.

Supported versions

Only the latest tagged release and the main branch receive fixes.

There aren't any published security advisories