philterd/philter-diffuse repository overview

⁠Philter Diffuse

Philter Diffuse adds calibrated noise to PII counts, so aggregate statistics can be shared with a formal bound on what each release reveals about any one document. It tracks the privacy budget spent on each source. It reads the counts Philter⁠ records, or a JSON file of counts.

The bound applies to one document, not one person. Philter Diffuse is early-access software.

Source and documentation: github.com/philterd/philter-diffuse⁠

⁠Quick start

mkdir data
echo '{"ssn": 105, "email-address": 450}' > data/pii_counts.json

docker run --rm --user "$(id -u):$(id -g)" -v "$(pwd)/data:/data" \
  philterd/philter-diffuse:0.1.0 \
  --input pii_counts.json \
  --source claims-2026 \
  --pii-types ssn,email-address \
  --output privatized_counts.csv

Always mount /data. It holds the input, the output, and the privacy budget, which is otherwise lost when the container exits.

⁠License

Apache 2.0.