endorlabs.comarticleMalicious Code in Open Source: How to Detect It Before You Ship | Blog | Endor Labs
Endor Labs explains the two dominant classes of OSS malware attacks - dependency/name confusion (typosquatting, combo-squatting, brand-jacking) and attacks on legitimate packages/maintainers - and how the OpenSSF malware-analysis project sandboxes new PyPI/npm releases and reports confirmed malware to OSV. It walks through real cases it detected, including the chunked/split-across-files PyPI malware ttlo and gisi, and Whatfuscator which drops and runs a PyInstaller-built Windows executable, plus two false-positive suspicious-code flags (versioneer 0.29, azure-functions 1.18.0) caused by benign base64/zlib obfuscation.
TL;DR Endor Labs breaks down dependency/name-confusion vs. legitimate-package-compromise malware, with real PyPI case studies (ttlo, gisi, Whatfuscator) and two benign obfuscation false positives, before pitching its own scanner.
Organisations named here
- azure-functions
- gisi
- OpenSSF malware-analysis
- ttlo
- versioneer
- werkzeug
Technologies named here
- Endor Labs Open Source
- PyPI
- npm
- OpenSSF malware-analysis project
- versioneer@0.29
- OSV
- azure-functions@1.18.0
- PyPI Stats
- GitHub Dependabot
- PyInstaller
- Python
- Endor Labs
- Windows
Techniques named here
- Dependency confusion attacks
- Name confusion attacks
- Splitting malicious code across multiple files/packages
- Downloading and executing a remote payload
- Code obfuscation via Base64 encoding and compression
- Attacks on legitimate packages
Threat actors named here
- ttlo and gisi
- Whatfuscator
- bobikssf
- Malicious GitHub Dependabot impersonation campaign
Related in Supply Chain
