How can we abuse AI hallucinations to feed the already bloated software supply chain
summary
Claim: generative models hallucinate plausible package names and attackers register those slopsquatting packages on PyPI DockerHub and Hugging Face to push malware. Demo shows fabricating a package, registering it, and measuring downstream impact — a real case yields 30,000 downloads over three months after a hallucinated suggestion and registries see billions of downloads per hour and 10 trillion/year forecasts. Defense uses concrete tooling like the OSV scanner, registry fuzzers and CI checks that detect provenance and block malicious package installs.
AI-written summary · about
chapters (8)
- 0:00AI and supply chain risks
- 1:05AI security blind spot
- 6:45Type squatting explained
- 11:41Open infrastructure costs
- 15:11Recent npm incidents
- 17:40OpenSSF and OSV scanner
- 21:54Malware detection demo
- 24:47Summary and security steps
related talks
35mAI Package Hallucination: Spreading Malicious Packages Using Generative AI
16mHarnessing AI for Malevolence: A Hackers Tale
40mThe Mechanisms That Enable Misuse and What We Can Do About It
37mAI-Powered Slopsquatting: Is Your Software Supply Chain Compromised?
56mHidden Risks in Open-Source Code and AI Models
45mThe Dark Side of AI: The Hidden Supply Chain Risks in Open...
01h 01mExploiting the supply chain
37mSponsored Session: The Simple, Yet Lethal, Anatomy of a Software Supply Chain Attack