PackageSentinel: Detecting Malicious npm & PyPI Packages with Static Analysis and AI

Modern applications rely on hundreds of open-source dependencies, making package ecosystems like npm and PyPI an attractive target for software supply chain attacks. While existing tools are excellent at identifying known vulnerabilities, they often don't answer a different but equally important question: "Can I trust this package?"

In this session, I'll introduce PackageSentinel, an open-source package analysis framework designed to identify suspicious behavior in npm and PyPI packages before they become part of your software supply chain.

We'll walk through how the tool downloads and analyzes packages using lightweight static analysis techniques, including detecting install-time scripts, dangerous API usage, obfuscated code, suspicious network activity, and access to sensitive files. These signals are combined into a risk score that helps developers prioritize packages for further review.

The talk also explores where Large Language Models fit into this workflow—not as a replacement for traditional security analysis, but as a way to explain findings, summarize suspicious behavior, and make security reports easier for developers to understand.

Finally, we'll see a live demonstration of PackageSentinel analyzing both legitimate and known malicious packages from npm and PyPI, discuss the trade-offs of heuristic-based detection, and look at how such a tool could integrate into CI/CD pipelines to improve software supply chain security.

Audience

• Software engineers

• Open-source maintainers

• DevSecOps engineers

• Application security professionals

• Students interested in software supply chain security

• Understand how malicious packages target npm and PyPI ecosystems.

• Learn practical static analysis techniques for package inspection.

• See how heuristic-based risk scoring can identify suspicious packages.

• Understand where AI adds value—and where it doesn't—in security tooling.

• Learn how developers can integrate package analysis into their software supply chain workflows.