Taming ML’s Wild West: Practical Model Signing with Sigstore on Kaggle
summary
Practical steps for attesting model integrity: use the Python library model-signing (pip install model-signing) with Sigstore to create short‑lived OIDC-backed certificates (10 minute lifetime), sign a directory of model files rather than bespoke formats, push signatures plus transparency log inclusion proofs, and verify hashes and trainer identity (GitHub OIDC) on download. Demos include Kaggle and NVIDIA hub integrations, a Jupyter notebook provenance capture for cell execution, and a provenance chain that detects unexpected extra releases (40 signed vs 38 expected).
AI-written summary · about
chapters (14)
- 0:00Introduction to model signing
- 1:54Live attack demonstration
- 4:43Trust issues in model hubs
- 6:27Model signing requirements
- 7:09Signing a trained model
- 9:35Model signing 1.0 launch
- 10:40Sigstore integration explained
- 12:27Kaggle model signing demo
- 13:29Nvidia model hub integration
- 16:47Provenance and notebook attacks
- 19:16Integrity and provenance claims
- 20:41Future of model signing
- 22:06Getting involved in projects
- 23:11Q&A session
related talks
27mTaming the Wild West of ML: Practical Model Signing with Sigstore on Kaggle
25mFrom Model To Trust: Building Upon Tamper-proof ML Metadata Records
23mEnd-to-End Secure ML Development
19mModel transparency for AI/ML security
19mWe Know Security but How Do We Secure GenAI End-to-End?
21mML Model Signing: Cryptographically Paving the Way to Provenance in Machine Learning Models
21mEnd-to-end secure ML Development
23mTrust, Track, and Verify: Securing AI Pipelines End-to-End