fix: classify PyTorch binary code patterns as findings

[mldangelo-oai]

Summary

• emit active PyTorch binary code-pattern detections as warning findings instead of info-only checks
• make aggregate security/exit logic return exit 1 for raw binary eval-like payloads
• preserve benign float tensor-like .bin handling

Validation

• original probe before fix: severity info, exit 0
• original probe after fix: severity warning, exit 1
• PYTHONPATH=/private/tmp/modelaudit-c132 PROMPTFOO_DISABLE_TELEMETRY=1 /Users/mdangelo/code/modelaudit/.venv/bin/pytest tests/scanners/test_pytorch_binary_scanner.py -q: 21 passed, 2 skipped
• ruff format/check for touched files
• mypy modelaudit/scanners/pytorch_binary_scanner.py tests/scanners/test_pytorch_binary_scanner.py
• git diff --check

[mldangelo-oai]

@codex review

[github-actions]

Workflow run and artifacts

Performance Benchmarks

Compared 12 shared benchmarks with a regression threshold of 15%.
Status: 0 regressions, 0 improved, 12 stable, 0 new, 0 missing.
Aggregate shared-benchmark median: 579.78ms -> 580.07ms (+0.1%).

Workload	Benchmark	Target	Size	Files	Baseline	Current	Change	Status
clean-training-checkpoint	tests/benchmarks/test_picklescan_benchmarks.py::test_picklescan_clean_training_checkpoint	safe_large	278.2 KiB	1	13.29ms	13.85ms	+4.2%	stable
chunked-upload-stream	tests/benchmarks/test_picklescan_benchmarks.py::test_picklescan_chunked_upload_stream	chunked_stream	278.2 KiB	1	15.69ms	16.30ms	+3.9%	stable
suspicious-pickle-intake	tests/benchmarks/test_scan_benchmarks.py::test_scan_suspicious_pickle_intake	suspicious-intake	183.8 KiB	4	79.41ms	76.69ms	-3.4%	stable
nested-payload-review	tests/benchmarks/test_picklescan_benchmarks.py::test_picklescan_nested_payload_review[nested_base64]	nested_base64	98 B	1	319.1us	311.7us	-2.3%	stable
nested-payload-review	tests/benchmarks/test_picklescan_benchmarks.py::test_picklescan_nested_payload_review[nested_raw]	nested_raw	78 B	1	314.0us	307.5us	-2.1%	stable
direct-malicious-upload	tests/benchmarks/test_picklescan_benchmarks.py::test_picklescan_direct_malicious_upload	malicious_reduce	52 B	1	1.13ms	1.11ms	-1.5%	stable
single-checkpoint-preflight	tests/benchmarks/test_scan_benchmarks.py::test_scan_single_checkpoint_before_load	single_checkpoint.pkl	183.0 KiB	1	30.44ms	30.89ms	+1.5%	stable
nested-payload-review	tests/benchmarks/test_picklescan_benchmarks.py::test_picklescan_nested_payload_review[nested_hex]	nested_hex	130 B	1	320.3us	315.8us	-1.4%	stable
warm-cache-rescan	tests/benchmarks/test_scan_benchmarks.py::test_scan_warm_cached_repository_rescan	release-candidate	547.3 KiB	32	50.26ms	49.68ms	-1.2%	stable
mixed-model-repository	tests/benchmarks/test_scan_benchmarks.py::test_scan_release_candidate_repository	release-candidate	547.3 KiB	32	222.96ms	225.16ms	+1.0%	stable
padded-multi-stream-upload	tests/benchmarks/test_picklescan_benchmarks.py::test_picklescan_padded_multi_stream_upload	multi_stream_padded	4.1 KiB	1	1.18ms	1.17ms	-0.9%	stable
duplicate-heavy-registry	tests/benchmarks/test_scan_benchmarks.py::test_scan_duplicate_registry_snapshot	registry-snapshot	915.2 KiB	13	164.47ms	164.29ms	-0.1%	stable

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Breezy!

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".