DySec: A Machine Learning-based Dynamic Analysis for Detecting Malicious Packages in PyPI Ecosystem
Sk Tanzir Mehedi, Chadni Islam, Gowri Ramachandran, Raja Jurdak · 2025
Read DySec: A Machine Learning-based Dynamic Analysis for Detecting Malicious Packages in PyPI Ecosystem by Sk Tanzir Mehedi, Chadni Islam, Gowri Ramachandran, Raja Jurdak (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-00324.txt.
Description
Sk Tanzir Mehedi and co-authors build DySec in this 2025 paper, a behavior watcher that catches malicious Python packages as they install. Attackers hide bad code in the open PyPI store through tricks like look-alike names and delayed payloads. Static scans and metadata checks miss such moves. DySec instead traces live behavior with kernel and user probes during installation.
It records thirty-six signals, from system calls to network traffic and folder access. The team built a set of over fourteen thousand packages, half of them malicious traces, run in isolation. DySec scored about ninety-six percent accuracy with sub-second delay, cutting missed attacks sharply versus static methods. It even flagged eleven packages PyPI had passed; hand checks confirmed six, and four were removed after the team's report.
Catalog record
| CREATOR | Sk Tanzir Mehedi, Chadni Islam, Gowri Ramachandran, Raja Jurdak |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 64 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-00324