Deceptive Code: Lessons Learned from Analysing 1M+ OSS Packages for Malicious Code

Open Source package distributed with typosquatting and starjacking. Project X is compromised to deliver malware to its 100k+ user base. Sounds familiar? This talk will jump straight into open source package analysis for malicious “code” identification with real life metrics from analysing over 1 million OSS packages for malicious code.

We will share lessons learned while monitoring npm, PyPI and other package registry ecosystems for malicious packages. We will showcase real life malicious OSS packages that we have detected and eliminated from public repositories using static code analysis and dynamic analysis techniques. We will focus on unique techniques and heuristics that we used to detect them. We will end the talk with a quick demo of vet, our free and open source next-gen Software Composition Analysis tool that is integrated with malicious packages data feed to protect software development teams in near realtime.

Example malicious packages reported by us:

• https://osv.dev/vulnerability/MAL-2025-4339

• https://osv.dev/vulnerability/MAL-2025-4029

• https://osv.dev/vulnerability/MAL-2025-3867

• https://osv.dev/vulnerability/MAL-2025-4079

• https://osv.dev/vulnerability/MAL-2025-3660

• https://osv.dev/vulnerability/MAL-2025-3543

• https://osv.dev/vulnerability/MAL-2025-2227

• https://osv.dev/vulnerability/MAL-2025-3542

• https://osv.dev/vulnerability/MAL-2025-3541

• Malicious open source packages are real, here are the evidences & metrics

• Static code analysis for malicious OSS code identification with real world examples

• Dynamic analysis for malicious package identification with real world examples

• How to use "SafeDep vet", our open source project, to protect against malicious packages