Never a dill moment: Exploiting machine learning pickle files
blog.trailofbits.com | blog | #ai-security | #supply-chain | #deserialization | #python | #exploit | #machine-learning | #fickling | #pickle
Summary
Most ML models are Python pickles, and loading one can run arbitrary code. Trail of Bits releases Fickling, a pickle decompiler and bytecode rewriter, showing how attackers craft malicious models.
- Published
- Collected
--- id: 1819 title: "Never a dill moment: Exploiting machine learning pickle files" source_domain: blog.trailofbits.com source_url: "https://blog.trailofbits.com/2021/03/15/never-a-dill-moment-exploiting-machine-learning-pickle-files/" published_at: "2021-03-15T15:06:18Z" tags: ["ai-security", "supply-chain", "deserialization", "python", "exploit", "machine-learning", "fickling", "pickle"] phying_url: "https://phying.com/item/1819" lang: en --- # Never a dill moment: Exploiting machine learning pickle files ## Summary Most ML models are Python pickles, and loading one can run arbitrary code. Trail of Bits releases Fickling, a pickle decompiler and bytecode rewriter, showing how attackers craft malicious models. ## Intelligence Metadata - **Sources:** blog.trailofbits.com - **Published:** 2021-03-15T15:06:18Z - **Original:** https://blog.trailofbits.com/2021/03/15/never-a-dill-moment-exploiting-machine-learning-pickle-files/ - **Phying News:** https://phying.com/item/1819