This directory contains tools for tracking differential privacy budgets, available as part of the Google differential privacy library. Currently, it provides an implementation of Privacy Loss Distributions (PLDs) which can help compute an accurate estimate of the total ε, δ across multiple executions of differentially private aggregations. Our implementation currently supports Laplace mechanisms, Gaussian mechanisms and randomized response. More detailed definitions and references can be found in our supplementary pdf document.
We provide basic examples on how to use the library in example.cc.
For running the example using Bazel, you need to have Bazel installed. Once that is done, run:
bazel build :all bazel run :example
The current version of the library is not supported on Windows.