Differential privacy

Differential privacy (DP) is a wide-ranging paradigm for statistically-guaranteed data privacy. The Privacy Engineering Team at the Wikimedia Foundation is currently exploring various uses for DP at WMF. This page (and its subpages) intend to document project statuses, design decisions, and potential future directions.

A diagram of how global DP might work at WMF

Available datasets

edit
Differentially private dataset Date range / granularity Download Documentation
Pageviews Historical (pre-2017) Data download README
Pageviews Historical (9 Feb 2017 – 5 Feb 2023) Data download README
Pageviews Current (6 Feb 2023 – present) Data download README
Geoeditors Weekly Data download README
Geoeditors Monthly Data download README

Docs and educational materials

edit

Other materials relevant to differential privacy efforts at WMF, including statements of purpose, docs, and educational materials.

Docs

edit

External documentation

edit

Code repositories

edit

Educational materials

edit

For a broad introduction to differential privacy as a concept, reading this series of blog posts on the subject is strongly recommended. Damien Desfontaines takes the reader through most of the need-to-know concepts of differential privacy.

Other presentations/educational materials are available below:

Active DP projects

edit

A list of DP projects that are actively being worked on, along with relevant documentation.

Completed DP projects

edit

A list of DP projects that have been completed and are currently being monitored for continued performance.

Proposed DP projects

edit

A list of DP projects that may be worked on in the future.