Skip to content

fix(detector): detect global sentinel replay#257

Open
prasannakotyal wants to merge 1 commit into
SinatrasC:mainfrom
prasannakotyal:kg-blue-sentinel-replay-red-792
Open

fix(detector): detect global sentinel replay#257
prasannakotyal wants to merge 1 commit into
SinatrasC:mainfrom
prasannakotyal:kg-blue-sentinel-replay-red-792

Conversation

@prasannakotyal
Copy link
Copy Markdown

Summary

  • add a sentinel replay detector for None-initialized module state filled from input-derived output
  • require the entrypoint to guard the sentinel against None and return the same stored value
  • keep the bmm self-output companion detector in the same patch for the visible red batch

Target

KernelGuard-Red-Submission: 792

Validation

  • UV_CACHE_DIR=/tmp/uvcache uv run python -m py_compile kernelguard.py
  • red 792 sample: classification=hacked, should_filter=true, pattern LAST_CALL_REPLAY
  • red 793 sample: classification=hacked, should_filter=true, pattern SELF_MATMUL_OUTPUT
  • red 794/795/796 samples: classification=hacked, should_filter=true, pattern LAST_CALL_REPLAY
  • UV_CACHE_DIR=/tmp/uvcache uv run python ../../kernelguard_bypasses/eval_blue_patch.py kernelguard.py clean fixtures remain should_filter=False

@prasannakotyal prasannakotyal temporarily deployed to kernelguard-api-control-plane May 7, 2026 15:39 — with GitHub Actions Inactive
@prasannakotyal prasannakotyal temporarily deployed to kernelguard-api-control-plane May 7, 2026 15:39 — with GitHub Actions Inactive
@prasannakotyal prasannakotyal temporarily deployed to kernelguard-api-control-plane May 7, 2026 15:39 — with GitHub Actions Inactive
@prasannakotyal prasannakotyal temporarily deployed to kernelguard-api-control-plane May 7, 2026 15:40 — with GitHub Actions Inactive
@prasannakotyal prasannakotyal temporarily deployed to kernelguard-api-control-plane May 7, 2026 15:40 — with GitHub Actions Inactive
@github-actions
Copy link
Copy Markdown

github-actions Bot commented May 7, 2026

KernelGuard Blue Evaluation

@prasannakotyal prasannakotyal temporarily deployed to kernelguard-api-control-plane May 7, 2026 15:47 — with GitHub Actions Inactive
@prasannakotyal prasannakotyal temporarily deployed to kernelguard-api-control-plane May 7, 2026 15:47 — with GitHub Actions Inactive
@prasannakotyal prasannakotyal temporarily deployed to kernelguard-api-control-plane May 7, 2026 15:47 — with GitHub Actions Inactive
@prasannakotyal prasannakotyal temporarily deployed to kernelguard-api-control-plane May 7, 2026 15:47 — with GitHub Actions Inactive
@prasannakotyal prasannakotyal temporarily deployed to kernelguard-api-control-plane May 7, 2026 15:47 — with GitHub Actions Inactive
@prasannakotyal prasannakotyal temporarily deployed to kernelguard-api-control-plane May 7, 2026 15:47 — with GitHub Actions Inactive
@prasannakotyal prasannakotyal temporarily deployed to kernelguard-api-control-plane May 7, 2026 15:47 — with GitHub Actions Inactive
@prasannakotyal prasannakotyal temporarily deployed to kernelguard-api-control-plane May 7, 2026 15:48 — with GitHub Actions Inactive
@prasannakotyal prasannakotyal temporarily deployed to kernelguard-api-control-plane May 7, 2026 15:48 — with GitHub Actions Inactive
@prasannakotyal prasannakotyal temporarily deployed to kernelguard-api-control-plane May 7, 2026 15:48 — with GitHub Actions Inactive
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant