# Detecting Malicious Source Code in PyPI Packages with LLMs: Does RAG Come in Handy?

arXiv: https://arxiv.org/abs/2504.13769 · HTML: https://mlsanity.com/paper/2504.13769/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Privacy and security of RAG, one of 15 directions in Retrieval-augmented generation.
- Builds on 11 of its 35 references that are in the map.
- First posted on arXiv 18 Apr 2025.

## Placed in

- Retrieval-augmented generation / [Privacy and security of RAG](https://mlsanity.com/rag/privacy-and-security-of-rag/)

## Builds on (most cited first, 11 of 11)

- [Large Language Models for Software Engineering: Survey and Open Problems](https://mlsanity.com/paper/2310.03533/) (2023)
- [Self-planning Code Generation with Large Language Models](https://mlsanity.com/paper/2303.06689/) (2023)
- [A New Era in LLM Security: Exploring Security Concerns in Real-World LLM-based Systems](https://mlsanity.com/paper/2402.18649/) (2024)
- [Vul-RAG: Enhancing LLM-based Vulnerability Detection via Knowledge-level RAG](https://mlsanity.com/paper/2406.11147/) (2024)
- [Backstabber's Knife Collection: A Review of Open Source Software Supply Chain Attacks](https://mlsanity.com/paper/2005.09535/) (2020)
- [PenHeal: A Two-Stage LLM Framework for Automated Pentesting and Optimal Remediation](https://mlsanity.com/paper/2407.17788/) (2024)
- [An Empirical Study of Malicious Code In PyPI Ecosystem](https://mlsanity.com/paper/2309.11021/) (2023)
- [A Comparison of Static, Dynamic, and Hybrid Analysis for Malware Detection](https://mlsanity.com/paper/2203.09938/) (2022)
- [Do Machines and Humans Focus on Similar Code? Exploring Explainability of Large Language Models in Code Summarization](https://mlsanity.com/paper/2402.14182/) (2024)
- [LProtector: An LLM-driven Vulnerability Detection System](https://mlsanity.com/paper/2411.06493/) (2024)
- [Development of recommendation systems for software engineering: the CROSSMINER experience](https://mlsanity.com/paper/2103.06987/) (2021)

## Built on by (most cited first, 3 of 3)

- [Many Hands Make Light Work: An LLM-based Multi-Agent System for Detecting Malicious PyPI Packages](https://mlsanity.com/paper/2601.12148/) (2026)
- [One Detector Fits All: Robust and Adaptive Detection of Malicious Packages from PyPI to Enterprises](https://mlsanity.com/paper/2512.04338/) (2025)
- [MOLOT System Card: Malicious Operational Logic Observation Transformer](https://mlsanity.com/paper/2606.07792/) (2026)

## Abstract (arXiv)

Malicious software packages in open-source ecosystems, such as PyPI, pose growing security risks. Unlike traditional vulnerabilities, these packages are intentionally designed to deceive users, making detection challenging due to evolving attack methods and the lack of structured datasets. In this work, we empirically evaluate the effectiveness of Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), and few-shot learning for detecting malicious source code. We fine-tune LLMs on curated datasets and integrate YARA rules, GitHub Security Advisories, and malicious code snippets with the aim of enhancing classification accuracy. We came across a counterintuitive outcome: While RAG is expected to boost up the prediction performance, it fails in the performed evaluation, obtaining a mediocre accuracy. In contrast, few-shot learning is more effective as it significantly improves the detection of malicious code, achieving 97% accuracy and 95% balanced accuracy, outperforming traditional RAG approaches. Thus, future work should expand structured knowledge bases, refine retrieval models, and explore hybrid AI-driven cybersecurity solutions.