Skip to content

Security: teleworksai/acn-bench

SECURITY.md

Security

This is pre-release research code. Only main is supported.

  • Reporting. Use GitHub's private vulnerability reporting for this repository ("Security" → "Report a vulnerability"). If that button is not shown, open an issue that says only that you need a security contact, with no details, and the maintainer will open a private channel. Do not describe a suspected vulnerability in a public issue.
  • Secrets. API keys live in .env, which is gitignored, and will be read only behind the real-api feature (it arrives with T04; nothing reads them today). Never commit a key, a provider response containing message content, or a packet capture with payload (CON-21, TRC-42). Use low-limit, project-scoped keys. Keys are never CI secrets for pull_request runs, because those runs execute the PR's code. If a key leaks, rotate it first and report afterwards.
  • Logs, traces and bundles. Code must never log a request's Authorization value or a key, and provider error bodies can echo both, so they are not logged verbatim. Bundles under runs/ are reviewed for content before they are published; a run made with message content kept (TRC-42) is never published.
  • Measured traces. Files under scenarios/measured/ must carry timing, size and link-state fields only, with a provenance file: no payload, no identifiers (CON-21). Nothing checks this mechanically yet; it is a checklist item for the author, and for the reviewer when there is one, on every PR that adds one.
  • Dependencies. cargo deny check runs on every PR for the workspace: advisories, licences, banned crates, and unknown registries or git sources. Lab crates are outside the workspace and are not covered: they may use any dependency, and their build scripts run with your user's rights, so do not build an unfamiliar lab crate on a machine whose .env holds real keys.
  • Privileged tests. The netem tier needs root. In CI it runs only on the schedule, from main, never on a pull request's code.

There aren't any published security advisories