NeurIPS 2019poster644 citations
Differential Privacy Has Disparate Impact on Model Accuracy
Eugene Bagdasaryan, Omid Poursaeed, Vitaly Shmatikov
Abstract
Differential privacy (DP) is a popular mechanism for training machine learning models with bounded leakage about the presence of specific points in the training data. The cost of differential privacy is a reduction in the model's accuracy. We demonstrate that in the neural networks trained using differentially private stochastic gradient descent (DP-SGD), this cost is not borne equally: accuracy of DP models drops much more for the underrepresented classes and subgroups.
BibTeX
@inproceedings{NEURIPS2019_fc0de4e0,
author = {Bagdasaryan, Eugene and Poursaeed, Omid and Shmatikov, Vitaly},
booktitle = {Advances in Neural Information Processing Systems},
editor = {H. Wallach and H. Larochelle and A. Beygelzimer and F. d\textquotesingle Alch\'{e}-Buc and E. Fox and R. Garnett},
pages = {},
publisher = {Curran Associates, Inc.},
title = {Differential Privacy Has Disparate Impact on Model Accuracy},
url = {https://proceedings.neurips.cc/paper_files/paper/2019/file/fc0de4e0396fff257ea362983c2dda5a-Paper.pdf},
volume = {32},
year = {2019}
}