gate-house/docs/conformance/2026-09-02-whitehat-asm-fixture-calibrations.md
tegwick c0c25e7056 Finish GH-WP-0001 conformance loop
Assistant: codex
Assistant-Model: gpt-5.6-sol
Assistant-Session: 01a05e30-2884-71b0-98d7-7edd16ae737b
2026-09-02 15:49:25 +02:00

2.8 KiB

Conformance review — Whitehat ASM fixture calibrations

Repository: gate-house Status: fixture harness calibrated; live targets pending Date: 2026-09-02 Executor: whitehat-security Executor revision: 6f1ca0b Source: State Hub message 32926191-4ca1-46ad-8eec-0d499036d66b Specification: asm-assurance-targets.v1 Evidence class: fixture Gate House disposition: no_change Workplan: GH-WP-0001-T06

Returned calibration

Whitehat-security created in-process registrations and evidence for Canon T-01 through T-10:

targets/fixture-asm-tNN.json
evidence/offline-asm-tNN-calibration.json

For every test, the Whitehat-owned known-bad fixture loses the specified oracle and records a finding; the corresponding known-good fixture records pass. T-08 and T-09 each calibrate two distinct failure shapes. T-06 drives the committed secrets-engine consume client at revision 4b4d556 through an in-process CAS opener; the other fixtures are Whitehat-owned in-process models of their target invariant.

Gate House reproduced the focused suite against Whitehat revision 6f1ca0b:

uv run --project /home/worsch/whitehat-security \
  pytest -p no:cacheprovider tests/test_asm.py tests/test_plane.py tests/test_cli.py
53 passed in 0.34s

This establishes that each published target has a probe capable of detecting a known-bad case. It closes the risk that a probe could report green merely because it cannot observe its own oracle.

Evidence bound

This is harness calibration, not assurance of ten estate systems:

  • evidence class is fixture;
  • no network, live service, OpenBao instance, credential, packet, or production effect was used;
  • no live asm-tNN registration became applicable;
  • no component other than the bounded T-06 consume client was exercised;
  • a fixture pass means the probe distinguished its known-good model from its known-bad model, not that the corresponding estate control currently holds.

All live T-01 through T-10 targets remain pending / not_run until an owner names the surface, identities, and window and Whitehat admits a dated engagement. WHITEHAT-WP-0007-T11 preserves that residual.

Doctrine disposition

no_change to asm-assurance-targets.v1.

The returned fixture set implements Whitehat-owned attack designs while preserving Gate House's invariant and oracle identifiers. No calibration exposed an ambiguous, contradictory, or untestable Gate House target. This disposition accepts the harness/specification fit only; it does not convert fixture outcomes into live conformance.

Reporting closure

Whitehat supplied the complete T-06 envelope, including engagement and authorization references, in message a1ebf012-bd1e-43d2-843c-ba3ddecb8c82. Final review: docs/conformance/2026-09-02-whitehat-t06-fixture-return.md.