Three Tools for Practical Differential Privacy
Machine Learning
2018-12-10 v1 Machine Learning
Abstract
Differentially private learning on real-world data poses challenges for standard machine learning practice: privacy guarantees are difficult to interpret, hyperparameter tuning on private data reduces the privacy budget, and ad-hoc privacy attacks are often required to test model privacy. We introduce three tools to make differentially private machine learning more practical: (1) simple sanity checks which can be carried out in a centralized manner before training, (2) an adaptive clipping bound to reduce the effective number of tuneable privacy parameters, and (3) we show that large-batch training improves model performance.
Cite
@article{arxiv.1812.02890,
title = {Three Tools for Practical Differential Privacy},
author = {Koen Lennart van der Veen and Ruben Seggers and Peter Bloem and Giorgio Patrini},
journal= {arXiv preprint arXiv:1812.02890},
year = {2018}
}
Comments
4 pages, 8 figures, PPML18: Privacy Preserving Machine Learning - NIPS 2018 Workshop