PyDP

PyDP is an open-source Python library for differential privacy, created by OpenMined in 2020. Machine learning models often handle sensitive or personal data, and some models can memorise details of their training data and later leak them. Differential privacy is a mathematical framework that measures and limits such leakage. It works by adding carefully calibrated random noise to results, so that no single person's data can be identified. PyDP makes this framework accessible to Python users, such as data scientists and researchers.

PyDP is a Python wrapper for Google's Differential Privacy library, which is written in C++. It gives Python users access to Google's tested algorithms without needing to work in C++. The library provides differentially private algorithms for producing summary statistics from numerical datasets that contain sensitive information. These include bounded mean, bounded sum, count, minimum, maximum, median and percentiles. All algorithms currently use Laplace noise.

For each calculation, users set a privacy parameter, known as epsilon, which controls the trade-off between privacy and accuracy. Lower values give stronger privacy but less accurate results. Users can also set upper and lower limits on the data values, or let PyDP calculate these automatically. PyDP runs on Linux, macOS and Windows. Tutorials and example notebooks help new users get started, and safe use follows the attack model of the underlying Google library. PyDP is released under the Apache 2.0 licence.

Auto-discovered on 2026-08-05 by OECD Catalogue Automation