data-privacy-stack/presidio

An open-source framework for detecting, redacting, masking, and anonymizing sensitive data (PII) across text, images, and structured data. Supports NLP, pattern matching, and customizable pipelines.

What it solves

Presidio provides a way for organizations to manage and govern sensitive data by identifying and anonymizing Personally Identifiable Information (PII) in text and images. It helps preserve privacy by democratizing de-identification technologies and making the process more transparent.

How it works

Presidio uses a pluggable and customizable architecture to detect and redact sensitive data. It employs several mechanisms for identification, including Named Entity Recognition (NER), regular expressions, rule-based logic, and checksums. It can also connect to external PII detection models. For images, it includes a specific module for redacting PII text within standard image types and DICOM medical images.

Who it’s for

It is designed for organizations that need to automate or semi-automate PII de-identification flows across multiple platforms, including Python, PySpark, Docker, and Kubernetes.

Highlights

  • Multi-modal support: Handles PII in both unstructured text and images (including medical DICOM images).
  • Flexible detection: Combines predefined and custom recognizers using NER, regex, and rules.
  • Extensible: Supports connecting to external PII detection models and is highly customizable to specific business needs.
  • Deployment flexibility: Can be deployed via pip, Docker, or Kubernetes.

Related

  • Project
    data-privacy-stack/presidio-research

    A toolkit for evaluating PII detection models and generating synthetic PII datasets using templates to improve Named Entity Recognition (NER) accuracy.

  • Dispatch
    Hugging Face PII Detection Experiment with Presidio

    Hugging Face is experimenting with integrating Microsoft Presidio into the Dataset Hub to provide automatic PII detection reports, helping practitioners identify and mitigate privacy risks in ML datasets.

  • Project
    Rizzo-AI-Academy/rizzo-pii

    A local, reversible PII anonymization tool for Italian legal text that allows users to use cloud LLMs without sending sensitive personal data.

  • Dispatch
    OpenAI Privacy Filter: Building Scalable PII Detection Web Apps

    OpenAI has released Privacy Filter, an open-source 1.5B-parameter PII detector capable of labeling eight categories of sensitive data across a 128k context window.

  • Dispatch
    OpenAI Privacy Filter Release

    OpenAI has released Privacy Filter, an open-weight, 1.5B parameter model designed for high-throughput, context-aware detection and redaction of personally identifiable information (PII) in unstructured text.