A free, open-source framework is the first thing many teams try when they need personal details taken out of documents. It costs nothing to license. What it asks for is engineering: someone has to install it, host it, tune it and build the part where a person checks the result.
What Presidio is
Presidio is an open-source framework for finding and anonymizing personal data in text, images and structured data. It began at Microsoft and is now an independent, community-governed project under the MIT licence. You run it yourself: a Python install with a language model, or a container.
Detection is a set of recognizers: patterns with context words, named-entity models, and any recognizer you write. Out of the box it works in English, and other languages need models and recognizers you add. Images are a separate module that needs an OCR engine you install, and the project describes that module as still in beta and not production ready.
What the project says about itself
Presidio is candid about its limits. Its own documentation says there is no guarantee that it will find all sensitive information, and that other systems and protections should be used with it. That is an honest notice, and it describes the work a deployment still has to do: measuring misses on your own documents, and putting a person in front of the result.
Side by side
| Question | Presidio | PHI Mask |
|---|---|---|
| What it is | An open-source framework you build with | A workspace for one document, and a service for a whole document set |
| Cost | Free to license; your engineers' time to run | A free plan for single documents; a batch is priced by the page |
| Scans and images | A separate module in beta, with an OCR engine you install | Read page by page, built in |
| A person checks the result | Yours to build | Each value is marked for you to approve; a batch has a reviewed sample |
| Where it runs | Wherever you host it | The browser tab you opened; a batch runs offline in your environment |
| Before the first file | An install, a model and code that calls it | Open a link |
Where Presidio is the better choice
If you have engineers and want full control, Presidio is hard to beat. It is free, the code is open, and you can write a recognizer for any identifier your organization uses. It fits inside a data pipeline or in front of a model, which is not where PHI Mask sits.
Where PHI Mask fits
PHI Mask is for the team that needs the documents done, not a framework to maintain. One document is masked in the browser. A whole document set goes through PHI Mask Batch, which runs offline in your environment or with our team, reads scans as it goes, and returns redacted copies with a release report and a sample that people have read. Detection results by document type are published on the benchmark.
Review before sharing.
Sources and review notes
Every statement about Presidio above comes from the project's own current documentation, reviewed on the date shown at the top of this page: its home page, the project transition note, installation, the analyzer, languages and the image redactor. Open-source projects change quickly, which is why this comparison carries a recheck date.
Common questions
- Is Presidio free?
- Yes. It is open source under the MIT licence. The cost is the engineering to install, host and tune it, and to build the step where a person checks the result.
- Is Presidio still a Microsoft project?
- It began at Microsoft. The project's own transition note describes it as an independent, community-governed open source project.
- Can Presidio read scanned pages?
- Through a separate image module with an OCR engine you install. The project describes that module as still in beta and not production ready.
- When is PHI Mask the wrong choice?
- When you want a framework to build into your own pipeline, with recognizers you write yourself. That is what Presidio is for.