Skip to content
Registry StackDocsv0.25.0

Security self-assessment

View as Markdown

This is a CNCF TAG Security-style self-assessment. Registry Stack is not currently a CNCF project, is not in the CNCF donation track, and has not been reviewed by CNCF TAG Security. This document is maintainer self-assessment against that public guide.

FieldValue
ProjectRegistry Stack
Repositoryregistrystack/registry-stack
Assessment typeMaintainer self-assessment
Review historyNo public review ledger yet
Last reviewed2026-08-03

Registry Stack helps operators publish governed registry-facing services over data they already hold. The two main runtime patterns are protected registry APIs through Registry Relay and signed minimum-disclosure assertions through Evidence Gateway.

The public architecture, boundary model, and product roles are documented in Architecture and Boundaries and map. The cross-cutting security requirements are documented in RS-SEC-G.

This assessment helps reviewers find public evidence quickly. It is not a certification, penetration test, production deployment proof, or CNCF review. The hosted lab is a synthetic-data behavior demo and does not prove production operational assurance.

Registry Stack security behavior is split between software guarantees and operator responsibilities.

Software guarantees include:

  • OIDC token validation through shared platform primitives.
  • Request-scope authorization and policy checks before protected reads or governed evidence evaluation.
  • Disclosure controls that can return narrower evidence instead of source records.
  • Signed, stateless Evidence Gateway assertions, serialized as a flattened JWS or, under Evidence Gateway’s frozen profile, as an SD-JWT VC response.
  • Hash-chained audit records emitted by runtime services.
  • Release workflow support for an authenticated checksum chain, exact image digests, a consolidated SPDX SBOM, image-specific SBOM and Grype evidence, and attestations for the exact candidate manifest and bundle.

Operator responsibilities include:

  • choosing and protecting signing-key storage;
  • configuring trusted issuers, audiences, scopes, and peer relationships;
  • retaining audit records in storage appropriate for the deployment’s integrity and retention needs;
  • reviewing vulnerability scan results and deployment-specific exposure before production use.

Registry Stack documents standards alignment in Standards and the public spec register in Specifications. These pages describe implementation evidence and known gaps; they do not claim legal compliance by themselves.

Development evidence:

  • CI runs Rust format, workspace checks, workspace tests, release-tool tests, lab source-model checks, and docs checks.
  • OpenSSF Scorecard runs on its schedule or by manual dispatch and publishes posture results as hygiene signals.
  • Release manifests pin source refs and external inputs.
  • Release validation checks source tag matching, source-ref ancestry, default branch reachability, and manifest consistency.
  • Current GitHub Release payloads are covered by one keyless Sigstore-signed SHA256SUMS file. The protected-main publication workflow verifies the candidate attestations before promotion. The public verification process is documented in release/VERIFY.md.

Known release-integrity gaps:

  • OCI image signatures are not yet published for root monorepo releases.
  • Ordinary Beta releases do not publish a second generic SLSA provenance asset. Candidate attestations and the signed checksum closure cover the current Beta provenance boundary.
  • Public claims about signing are scoped to the release asset or product surface that actually performs signing.

Report vulnerabilities through SECURITY.md. The project publishes public review-ledger entries only when they are fixed and public, or when an open theme is already publicly inferable and approved for publication by the security reviewer.

Primary public evidence: