habanerocake/pii-filter

A personally identifiable information (PII) filter.

Filter

Please note: this project is still experimental

A library for detecting, parsing, and removing personally identifiable information from strings and objects.

---

This repository contains the datasets and the npm module for pii-filter.

A showcase example of this library is the pii-filter web-extension for browsers.

Scenarios

We hope that this software can be useful in some of the following scenarios:

• privacy, security, fraud-detection, and data-auditing
• anonymizing data for research, marketing and machine learning
• accessibility and online guidance
• word tagging and word spotting
• designing chatbots

Languages

pii-filter currently supports the following languages and PII:

• Dutch
  • First Names
  • Family Names
  • Pet Names
  • Medicine Names
  • Phone Numbers
  • Email Addresses
  • Dates

More information can be found in dataset/README.md.

Future work

• adding support for english and other languages
• detecting more PII
  • bank numbers and credit card numbers
  • passport / id numbers
  • social security numbers
• detecting PII in images
  • providing coordinates which can be used for blurring or erasing PII

pii-filter's Issues

Better Classification

Classification quality can be improved drastically by using better models, if we can get access to labeled training data.

False Positives

Various false positives pop up if a phrase is not determined to be well-formed. This functionality should be removed or disabled until better classifiers are implemented.