v0.1.0 - Initial release

First public release of pdf-defang — a lightweight Python library for stripping JavaScript, OpenAction, Launch actions, and other active content from PDFs.

Install

pip install pdf-defang

Highlights

• Two levels: strict (default) removes all active content; balanced preserves form interactivity and embedded files for trusted sources.
• 6 public APIs: sanitize / scan, plus async variants for FastAPI/aiohttp, plus bytes variants for S3/Lambda pipelines.
• 9 categories of dangerous content stripped: document JavaScript, /OpenAction, document/page/annotation /AA, XFA forms, embedded files, dangerous annotation actions (/Launch, /GoToR, /ImportData, etc.), annotation /JS keys, and unsafe URI schemes (javascript:, file:, data:, UNC paths).
• Full encryption preservation when a password is supplied.
• CLI: pdf-defang clean / pdf-defang scan with --level flag.
• Field-tested against 4,558 real-world PDFs (1.96 GB), zero page-count regressions.

Quality

• 133 automated tests
• 90% code coverage
• mypy --strict clean (6 files)
• ruff clean
• CI matrix: Python 3.9 - 3.13 × Linux, macOS, Windows

Quick start

from pdf_defang import sanitize

# Public uploads (safest default):
sanitize("untrusted.pdf")

# Trusted internal form that needs Submit/Calculate buttons:
sanitize("expense_form.pdf", level="balanced")

Origin

Extracted from production code at kovetz.co.il after an APT scanning campaign targeted PDF upload endpoints in May 2026. See the origin story for the full background.

Documentation

• Full docs site: https://kovetz-pdf.github.io/pdf-defang/
• API reference: https://kovetz-pdf.github.io/pdf-defang/api/
• What gets removed: https://kovetz-pdf.github.io/pdf-defang/protections/
• Comparison with Dangerzone: https://kovetz-pdf.github.io/pdf-defang/compare/dangerzone/

License

MIT.