Phinder

A Java application that uses Phileas to identify PII (Personally Identifiable Information) in text across a wide variety of file formats. Types of PII are scored by risk, magnitude, density, and confidence, so the most sensitive findings can be prioritized. A list of files suggested for redaction testing will be generated.

The goal of Phinder is to provide a comprehensive analysis of PII to help you take the next step to redact it with Philter. Note that Phinder may support more file types than Philter.

Visit http://philterd.github.io/phinder for documentation and more information.

Have questions or want to get involved? Users and developers are welcome to join the Philterd Slack workspace.

Example Generated Report

Quick Start

Build the project

mvn clean install

Run Phinder

java -jar target/phinder-1.0.0-SNAPSHOT.jar -i src/test/resources/input.txt

To process a directory recursively:

java -jar target/phinder-1.0.0-SNAPSHOT.jar -i src/test/resources/ -R

Note

Processing images requires tesseract-ocr to be installed.

At the completion of the scan, report.json and report.html files will be generated in the current directory.

Prioritize by risk

Phinder scores every file by how sensitive the PII in it is, not just how much of it there is, so you can decide what to remediate first. The Risk Score combines the severity of each entity type, how many occurrences were found, and how confident the detector was, and every score is reported with a risk level from NONE to CRITICAL:

java -jar target/phinder-1.0.0-SNAPSHOT.jar -i src/test/resources/ -R --sort-by risk --min-risk 50 --top 10

A file with two Social Security numbers ranks above a file with fifty city names. Severities ship with a built-in default per entity type and can be tuned with --severities; see the risk scoring documentation for the table and the formula. The score is a way to order your work, not a compliance determination: detection is probabilistic, and you are responsible for reviewing the findings against your own data.

Generate a starter redaction policy

Phinder can turn a scan into a starter Philter / Phileas redaction policy that enables the entity types it found, so discovery and redaction become one workflow. Use --emit-policy with an output file:

java -jar target/phinder-1.0.0-SNAPSHOT.jar -i src/test/resources/ -R --emit-policy starter-policy.json

The output is redaction-policy JSON that loads unchanged into Philter or Phileas. Each detected type is enabled with a REDACT strategy. Apply it directly, or tune it first (change strategies, add conditions, ignore terms) and re-run. Use a .phisql file extension to get the policy as PhiSQL instead, which compiles to the same JSON.

This is a starting point to review, tune, and measure (for example with Philter Scope) before you rely on it. Redaction is probabilistic: it reduces how much sensitive data gets through, it does not catch every instance, and you are responsible for validating the policy against your own data. Types that need a supplied custom policy to detect (custom identifiers, sections, or PhEye name detection) are reported as skipped rather than guessed at.

Store report history in MongoDB

To store the report history in MongoDB, use the --mongodb CLI option:

java -jar target/phinder-1.0.0-SNAPSHOT.jar -i src/test/resources/input.txt --mongodb "mongodb://localhost:27017/phinder"

For more examples and detailed usage, please refer to the documentation.

License

Copyright 2026 Philterd, LLC.

This project is licensed under the Apache License 2.0.