Governance & compliance

PII detection, redaction and classification

Would find and mask personal data such as email addresses, phone numbers and card numbers before it reaches a model or leaves in an export.

RoadmapEngine edition 5f8ea3af

What it does

Not reachable in this release. The detection and redaction code exists as a library with tests, but no server route, command or SDK call uses it, so it does not run on any data in a deployment.

How it works

The library has deterministic detectors for email addresses, phone numbers, social-security-number-shaped values, card numbers with a Luhn check, and IPv4 addresses. A redaction is either irreversible, which replaces the value with a placeholder, or reversible, which substitutes a token and keeps an encrypted token map under a key referenced by id. Enforcement points would redact a model request or an export. A review step lets a person confirm or reject a detected span, and an evaluation module measures false positives and false negatives on a labelled set. A separate classification module scores findings continuously.

The server does not depend on this library, so no route can create a policy or run a redaction. Its documentation describes the detectors as heuristic and the whole as a reviewable mechanism, not a compliance certification.

Limits

  • Not reachable by users in this release.
  • The detectors are heuristic and report measurable false positives and false negatives.

Evidence

Test coverage is listed as present in the engine repository at this edition. This page does not report test results.

KynetraDB is in private preview

This is not available in the preview. Request a preview to discuss the current scope.