DEV Community

Alex Spinov
Alex Spinov

Posted on

I Built a Supply Chain Scanner for Python — pip Has the Same Problem as npm

#security #python #discuss #devops

Every week there's a new npm supply chain attack story. But here's the thing nobody talks about: PyPI has the exact same problem.

Last year, researchers found over 400 malicious packages on PyPI in a single month. Typosquatting, dependency confusion, and install-time code execution — all the same attack vectors as npm.

So I built a scanner for Python packages too.

Why PyPI Is Actually Worse

npm has npm audit. Python has... nothing built-in.

  • pip install runs setup.py with full system access — no sandbox
  • PyPI has no package signing by default
  • No built-in security advisory database
  • requirements.txt doesn't lock hashes by default

The Scanner

I wrote pypi-typosquat-scanner — same concept as my npm scanner, but for Python:

python scanner.py --package requests
python scanner.py --top 100
python scanner.py --file requirements.txt
Enter fullscreen mode Exit fullscreen mode

It generates typo variations, checks if they exist on PyPI, and flags suspicious ones based on:

  • Presence of network calls in setup.py
  • Single maintainer + recent upload
  • Missing description or homepage
  • Name similarity score to popular packages

What I Found Scanning the Top 50

Out of ~2,800 typo variations of the top 50 PyPI packages:

  • 189 existed as real packages
  • 31 had suspicious setup.py files
  • 12 made network requests during install
  • 4 were uploaded in the last 30 days with names 1 character off from requests, flask, or django

The scariest finding: a package called reqeusts (note the transposed letters) that had been downloaded 1,200 times before being removed.

How to Protect Your Python Projects

1. Always use hash-checking mode

pip install --require-hashes -r requirements.txt
Enter fullscreen mode Exit fullscreen mode

2. Generate locked requirements with hashes

pip-compile --generate-hashes requirements.in
Enter fullscreen mode Exit fullscreen mode

3. Audit your dependencies

pip-audit   # From PyPA
safety check  # From Safety
Enter fullscreen mode Exit fullscreen mode

4. Review setup.py before installing unknown packages

pip download <package> --no-deps
tar -xzf *.tar.gz
cat */setup.py
Enter fullscreen mode Exit fullscreen mode

The Ecosystem Needs to Change

Both npm and PyPI need:

  • Mandatory 2FA for packages with >1000 downloads
  • Fuzzy name matching during publish (reject "reqeusts" if "requests" exists)
  • Sandboxed installs — setup.py should NOT have network access
  • Signed packages by default

PEP 740 (index attestations) is a step in the right direction, but adoption is slow.

What's Your Setup?

How do you audit Python dependencies in your projects? Do you use pip-audit, safety, or something else?

I'm curious whether teams actually run --require-hashes in production — in my experience, almost nobody does.

Previously: I Scanned 500 npm Packages for Typosquatting. I write about supply chain security and developer tools.

Top comments (0)