Skip to content

Back to the catalog

Presidio (Microsoft)
github.comrepo

Presidio (Microsoft)

Presidio is Microsoft's open-source PII de-identification service that detects and anonymizes sensitive entities (credit card numbers, SSNs, names, locations, bitcoin wallets, phone numbers) in text and images, including DICOM medical images. It combines Named Entity Recognition, regex, rule-based logic, and checksum validation, and is deployable via Python, PySpark, Docker, or Kubernetes with pluggable custom recognizers.

TL;DR Microsoft's open-source toolkit for detecting and redacting PII from text and images.
Open original ↗
Document

What it is

Presidio is Microsoft's open-source PII detection and de-identification service for text and images. The page opens with a notice that the project "has moved to a new home". The README does not give the new location, and the contact address is now presidio@dataprivacystack.org.

Preconditions / Affected

  • Anyone pinning to github.com/microsoft/presidio should check where releases and issues now live before depending on that repository. The fetched text does not say whether the Microsoft repository is archived or still updated.
  • The README says recognizers work "in multiple languages" and that Presidio has "options for connecting to external PII detection models". It names no languages or supported models.
  • Image redaction covers standard image types and DICOM medical images.

What to do

  • Follow the move notice to the new home and confirm the current release source and issue tracker.
  • Per the README, read the documentation first, use GitHub Discussions for general questions, and file a GitHub issue for usage questions, bugs or suggestions.
  • Treat Presidio as one layer of protection, not the only one (see Caveats).

Caveats

The README states that detection is automated, so "there is no guarantee that Presidio will find all sensitive information". It says additional systems and protections should be used alongside it. The page gives no accuracy figures, benchmarks, version numbers or install commands.

Technologies named here

Techniques named here

  • PII identification and de-identification
  • PII redaction in images

Related in other