An advanced Differential Privacy library for pyTorch, based on opacus.

Jens Keim a39ff65a6a allow lists or tuples for weights as well %!s(int64=3) %!d(string=hai) anos
LICENSE-2.0.txt cf57afea64 add license %!s(int64=3) %!d(string=hai) anos
LICENSE.md cf57afea64 add license %!s(int64=3) %!d(string=hai) anos
README.md a88815e07c fix readme formatting for gogs %!s(int64=3) %!d(string=hai) anos
privacy_engine_xl.py a39ff65a6a allow lists or tuples for weights as well %!s(int64=3) %!d(string=hai) anos

README.md

aDPtorch

An advanced Differential Privacy library for pyTorch, based on opacus.

Current Features

  • privacy_engine_xl
    • additional noise distributions
      • Gauss (Normal)
      • Laplace
      • Exponential
    • server side noise generation function

License

This repository is licensed under the Apache License 2.0, as some of the code is derived from opacus.

This repo contains a markdown and a text version of the license.

In case of any inconstancies refer to the license's website.