DITTO: A Context-aware Pickle-based Pre-Trained Model Scanner for Effective Security Audits
DITTO introduces context-aware scanning for Pickle-serialised ML models, targeting deserialization exploits that conventional tools miss — directly relevant as AI supply chains scale in enterprise environments.
Summary written by editorial AI · Source link below
arXiv:2610.10735v1 Announce Type: new Abstract: Pre-trained models (PTMs) are widely distributed as serialized binaries, but their reuse often exposes software supply chains to deserialization attacks. Despite the emergence of safer serialization formats, the unsafe Pickle format remains prevalent: our analysis of over 10,000 popular Hugging Face repositories reveals that 9.3% rely on Pickle. While many defense mechanisms have been proposed, state-of-the-art model scanners suffer from a coverag
Editorial Analysis
With Pickle remaining the dominant ML model serialisation format despite known risks, automated detection of embedded malicious code is critical for securing AI supply chains.
Add Pickle deserialization scanning to your ML model ingestion pipeline, especially for externally sourced pre-trained models.
ML models distributed as Pickle files can carry hidden code — new tooling detects these supply-chain threats before deployment.
Forward-looking interpretation drafted by editorial AI under human review — not a reproduction of the source. See methodology.
External link — opens at arXiv Crypto & Security in a new tab.
More from the DevSecOps Desk
- 11 of 23 Core Open Source Projects Run on 1 or 2 People9 Oct
- Credential-Stealing GitHub Actions Workflows Planted in Tens of Thousands of Repositories9 Oct
- PackMonitor: Enabling Zero Package Hallucinations Through Decoding-Time Monitoring9 Oct
- Applying Security by Design at the Point of Execution: How Governed Security Requirements Affect the Security of AI-Generated Code9 Oct
- Arbiter: Detecting Interference in LLM Agent System Prompts9 Oct