AuraBorealis is a web application for visualizing anomalous and potentially malicious code in Python package registries. It uses security audit data produced by scanning the Python Package Index (PyPI) via Aura, a static analysis designed for large scale security auditing of Python packages. The current tool is a proof-of-concept, and includes some live Aura data, as well as some mockup data for demo purposes.

Current features include:

• Scanning the entire python package registry to:

  • List packages with the highest number of security warnings, sorted by Aura warning type
  • List packages sorted by the total and unique count of warnings
  • List packages by their overall severity score
  • List packages that have changes in their warnings and/or severity score between two dates

• Displaying security warnings for an individual package, sorted by criticality

• Visualize the line numbers and lines of code in files generating security warnings for a specific package

• Compare two packages or different versions of the same package for security warnings

Clone the repository.

git clone https://github.com/IQTLabs/AuraBorealisApp.git

Navigate to aura-borealis-flask-app directory.

cd aura-borealis-flask-app

Install dependencies.

pip install requirements.txt

Run the app.

python app.py

Navigate to the URL http://0.0.0.0:5000/ via a browser.

• Compare a package to a benchmark profile of packages of similar purpose for security warnings
• Ability to scan an internal package/registry that's not public on PyPI
• Display an analysis of permissions (does this package make a network connection? Does this package require OS-level library permissions?)

[email protected] (John Speed Meyers, Secure Code Reuse project lead). The lead developer and creator of Aura is Martin Carnogusky of sourcecode.ai.

• IQT blog post on secure code reuse
• IQT blog posts on typosquatting and preventing typosquatting via pypi-scan
• USENIX article on "Counting Broken Links: A Quant's View of Software Supply Chain Security"
• IQT open source dataset on known software supply chain compromises