Skip to content

v0.3.0 — Verification Context v1.0 + Attestation Trust Boundary

Latest
Latest

Choose a tag to compare

View all tags
@rahuldass19 rahuldass19 released this 25 Aug 10:09
v0.3.0
b3098be
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

QWED-Infra v0.3.0 ships Verification Context v1.0 across all four guards and the Attestation Trust Boundary — every ADMIT is now backed by an ES256-signed receipt bound to the exact claim and evidence.

✅ Added

Verification Context v1.0 (tracker #36)

  • to_verification_context() on all four guards — IamGuard (#45), NetworkGuard (#46), CostGuard (#48), ArtifactBoundaryGuard (#50) — producing schema-valid, tamper-evident VC documents (claim, verifier identity, sha256-bound proof_ref, ADMIT/DENY admission)
  • Shared verification_context_bridge module (#37/#44): diagnostic → VC document conversion with fail-closed attestation policy and decision-status demotion
  • Conformance suite tests/test_vc_conformance.py (#42/#52) — bridge + all guards + document validation + malformed-input fail-closed acceptance tests
  • README "Verification Context v1.0" section: why VC, usage examples, downstream integration guide

Attestation trust boundary (#47/#54)

  • New qwed_infra/attestation.py: ES256 (ECDSA P-256) JWT attestation service — signing, verification (signature/issuer/expiry/revocation), revocation registry, ephemeral key lifecycle auditing, never-None fail-closed AttestationResult
  • enforce_trust_decision(): single consumption-side gate validating tokens plus full claim binding (status, query_hash == sha256(formal_statement), proof_hash == diagnostic proof_ref)
  • mint_diagnostic_attestation(): issue a token bound to a diagnostic's own evidence commitment

⚠️ Changed

  • BREAKING (pre-1.0 API): guard adapters no longer accept pre-computed result objects. to_verification_context() takes raw verification inputs and runs the guard's deterministic solver internally — a result-accepting signature was forgeable (fabricated results could mint ADMIT), so it was removed before release.
  • ADMISSION SEMANTICS: VERIFIED results admit only with a cryptographically valid attestation bound to the exact claim and evidence. Arbitrary non-empty attestation strings no longer grant ADMIT (rejected as forged); a missing token demotes VERIFIED to UNVERIFIABLE/DENY. Callers using placeholder tokens must mint via create_verification_attestation() / mint_diagnostic_attestation() instead.
  • New runtime dependencies: pyjwt>=2.8.0,!=2.12.1, cryptography>=41.0.0

🔒 Fixed

  • Fail-closed on malformed inputs at every VC boundary: undecimal budgets, extreme Decimal values, non-string build backends, malformed topology/policy/package inputs, symlink escapes/loops, wheel entries outside the scanned boundary — all map to BLOCKED/DENY documents instead of exceptions or guessed approval (#48, #49/#51, #50)
  • ArtifactBoundaryGuard proof binding: per-file sha256 content manifests, package identity derived from the inspected directory, exact-match build-backend allowlist
  • IAM verified-denial results demote to BLOCKED (never ADMIT), with the document-level proof_ref staying null for every DENY decision

📦 Upgrade notes

  • If you passed placeholder strings as attestation_token, either drop the argument (VERIFIED → UNVERIFIABLE/DENY) or mint real attestations bound to your claim+evidence
  • Guard adapter calls change shape: pass raw inputs directly, e.g. NetworkGuard().to_verification_context(infra, source, dst, port, formal_statement=..., attestation_token=...)
  • Build reproducibility: hatchling==1.27.0 pinned (Metadata-Version 2.4)

Full changelog: see CHANGELOG.md

Assets 2
Loading