Skip to content

Security: datapip/genug-analytics

SECURITY.md

Security policy

Genug collects data about other people's visitors, so a security hole here is a privacy problem for people who never chose to use it. Thank you for taking the time to report one.

Reporting a vulnerability

Please don't open a public issue. Report it privately through GitHub: Security → Report a vulnerability on this repository (direct link). Only the maintainer sees the report, and the fix can be prepared before anything is public.

If you can't use GitHub, write through the contact form. Say that it is a security report, but leave out the details, and a private channel will be set up.

A useful report says what an attacker can do, what they need first (a cockpit session, the MCP key, nothing at all), and how to reproduce it. A proof of concept against your own deployment is welcome. Please don't test against a deployment that isn't yours.

What to expect

Genug is maintained by one person in their spare time. You'll get a first reply within 7 days, and an estimate once the problem is understood. Fixes ship as a new version, with a line in CHANGELOG.md and a published advisory crediting you, if you want that.

Supported versions

Only the latest release gets fixes. There is no long-term branch: an older version is updated by moving to the newest one, which CHANGELOG.md describes.

What counts

Anything that breaks a promise Genug makes, for example:

  • /events (public, no sign-in) storing, revealing or executing something it shouldn't, or taking the server down with less effort than the documented rate limits allow.
  • Getting into the cockpit or /mcp without the password or key, or doing more than READ_ONLY allows.
  • Visitor-supplied text reaching the agent as an instruction rather than as data.
  • A visitor being identified more precisely than docs/privacy.md says, for example linking a consentless visitor across days.
  • Data that erasure or retention should remove surviving on the live database.

Out of scope: problems that need the server's own files or shell access, a deployment configured against docs/deploying.md (for example a proxied record without TRUST_PROXY), and missing hardening with no attack attached. Those are still welcome as ordinary issues.

There aren't any published security advisories