Malicious Package Detection on PyPI Large Packages N=1162 (test)
96.89AccuracyH2GLM
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| H2GLM2026.07 | 96.89 | 95.51 | 98.4 | 96.93 | |
| MPHunter2026.07 | 86.4 | 81.1 | 87.4 | 84.13 | |
| HERCULE2026.07 | 86.14 | 83.28 | 90.45 | 86.72 | |
| Qwen3-14b2026.07 | 77.92 | 96.84 | 57.72 | 72.33 | |
| Warehouse2026.07 | 63.94 | 89.32 | 31.67 | 46.76 | |
| OSSGadget2026.07 | 60.15 | 47.16 | 89.33 | 61.73 | |
| DeepSeek-R1-70b2026.07 | 57.25 | 60.08 | 43.22 | 50.27 | |
| Naive Bayes2026.07 | 51.98 | 52.02 | 51.03 | 51.52 | |
| Llama3-8b2026.07 | 43.08 | 45.01 | 62.43 | 52.31 | |
| Qwen2.5-7b2026.07 | 42.51 | 40.29 | 31.07 | 35.09 | |
| Bandit4Mal2026.07 | 41.2 | 35.8 | 18.2 | 24.13 | |
| Clustering2026.07 | 36.38 | 20.96 | 9.83 | 13.38 |