What is this?
An open-source framework (SDK) that detects, redacts, masks and anonymizes sensitive data (PII) — names, credit card numbers, SSNs, phone numbers — in text, images and structured data.
An open-source framework (SDK) that detects, redacts, masks and anonymizes sensitive data (PII) — names, credit card numbers, SSNs, phone numbers — in text, images and structured data.Is Presidio a library I import into my code, or a service I deploy?Both: it ships as Python packages you import in your own code, and the same components can also run as Docker containers or on Kubernetes.
Both: it ships as Python packages you import in your own code, and the same components can also run as Docker containers or on Kubernetes.How do I add it to my project?Install the Python packages with pip, run the provided Docker images, or build from source; the docs include a Getting started guide for both text and image de-identification.
Install the Python packages with pip, run the provided Docker images, or build from source; the docs include a Getting started guide for both text and image de-identification.Do I need to bring my own ML models, or is detection ready out of the box?Detection is powered by predefined or custom recognizers: NER (spaCy, Transformers), regular expressions, rule-based logic and checksums. External PII detection models can also be connected.
Detection is powered by predefined or custom recognizers: NER (spaCy, Transformers), regular expressions, rule-based logic and checksums. External PII detection models can also be connected.Can I rely on it to catch every piece of sensitive data?No — it relies on automated detection, so there is no guarantee it will find all sensitive information; the authors recommend employing additional systems and protections alongside it.
No — it relies on automated detection, so there is no guarantee it will find all sensitive information; the authors recommend employing additional systems and protections alongside it.What can it detect out of the box?Predefined recognizers cover names, locations, credit card numbers, social security numbers, US phone numbers, bitcoin wallets, financial data and more; custom recognizers can be added for domain-specific entities.
Predefined recognizers cover names, locations, credit card numbers, social security numbers, US phone numbers, bitcoin wallets, financial data and more; custom recognizers can be added for domain-specific entities.Which components does the project consist of?Four modules: Analyzer (finds PII), Anonymizer (redacts or masks it in text), Image Redactor (for standard images and DICOM medical images) and Presidio Structured (for structured data).
Four modules: Analyzer (finds PII), Anonymizer (redacts or masks it in text), Image Redactor (for standard images and DICOM medical images) and Presidio Structured (for structured data).Does it work with languages other than English?Yes — recognizers leverage NER, regex and rule-based logic with relevant context in multiple languages, so detection is not limited to English.
Yes — recognizers leverage NER, regex and rule-based logic with relevant context in multiple languages, so detection is not limited to English.What does the name Presidio mean?It comes from the Latin praesidium, meaning protection or garrison — a nod to its goal of guarding sensitive data.
It comes from the Latin praesidium, meaning protection or garrison — a nod to its goal of guarding sensitive data.What are similar projects?datamimic, spaCy, jevals, openmed, allennlp, privacy-preserving-control-plane, flair, CoreNLP
datamimic, spaCy, jevals, openmed, allennlp, privacy-preserving-control-plane, flair, CoreNLPWhich topics is it in?pii, nlp-library, medical-imaging, data-privacy, named-entity-recognition, healthcare, guardrails, spark, compliance, text-processing, privacy-tools, python-library
pii, nlp-library, medical-imaging, data-privacy, named-entity-recognition, healthcare, guardrails, spark, compliance, text-processing, privacy-tools, python-library