Smoke tests only — not a full production benchmark suite.
The runner signs results with HMAC-SHA256 so evidence is immutable and auditable.
Schema validation confirms result shape. Tests pass but coverage is shallow — each detector gets a handful of smoke-test cases, not comprehensive benchmark suites.
Smoke
Test depth
Not full benchmarks
HMAC
Result signing
SHA-256
✓
Schema validation
Result shape checked
7
Detectors covered
1–3 smoke tests each
Benchmark runs — evidence registry
| Detector |
Tests |
Status |
Signed evidence |
Run date |
| hf-model-provenance-scanner | 3 | Pass | evidence/hf_scan_smoke.json | 2026-08-01 |
| mcp-agent-security-gateway | 2 | Pass | evidence/mcp_smoke.json | 2026-08-01 |
| llm-redteam-framework | 3 | Pass | evidence/llm_smoke.json | 2026-07-30 |
| adversarial-ml-lab | 2 | Pass | evidence/adv_smoke.json | 2026-07-29 |
| dataset-poisoning-detector | 2 | Pass | evidence/poison_smoke.json | 2026-07-28 |
| model-privacy-attacks | 1 | Pass | evidence/privacy_smoke.json | 2026-07-27 |
| attack-detection-engine | 2 | Pass | evidence/engine_smoke.json | 2026-07-26 |