Sensitive actions leave an audit trail
The service records who did what to what, and when, for actions that matter in a dispute or an investigation.
MediumChecked by hand
- Applies to:
- Every service
- Rule id:
- security.audit-log
Why it matters
After a breach, a fraud claim or a customer complaint, the first question is who did what. Application logs rarely answer it: they are sampled, rotated and written for debugging. Without an audit trail the answer is a guess, and so is the report to the regulator or the customer.
How to apply it
- Decide which actions are sensitive: sign-in, permission changes, refunds, exports, reading or deleting personal data, admin overrides.
- For each, record the actor (user and calling service), the action, the target, the time, the result and the request id.
- Record failures and refusals too.
- Keep the trail apart from debug logs: append-only, with its own retention and restricted access.
- Keep personal data out of it; ids are enough.
- Alert on patterns that should not happen, such as an export at 3 a.m. by a support account.
Example
{"at":"2026-03-02T09:14:07Z","actor":"user:u_77","via":"support-console","action":"order.refund","target":"order:8841","result":"allowed","requestId":"4bf9…"}, written to an append-only store kept for seven years.
References
Checklists with this rule
Related agent skill: microservice-security