Published June 23, 2026 | Version v1

H2GLM: Malicious PyPI Package Dataset (2134 samples)

Authors/Creators

Description

This dataset contains 2134 unique malicious Python packages collected from PyPI, used for the evaluation of H2GLM (Hierarchical Heterogeneous Graph Learning framework enhanced by LLMs for Malicious package detection). Malicious samples are derived from a large-scale benchmark for malicious Python package detection, supplemented by packages collected from public security advisories and threat intelligence feeds. Deduplication was enforced via MD5 and fuzzy hashing. The archive malicious_2134.tar.gz contains the original .tar.gz distribution files as collected from PyPI. Intended for academic security research only.

Files

Files (156.0 MB)

Name Size Download all
md5:9fdf6993654704f62098148bd5672e57
156.0 MB Download