Improving Deep Learning with Differential Privacy using Gradient Encoding and Denoising

TL;DR

This paper introduces a novel method for training deep learning models with differential privacy that maintains higher accuracy by encoding gradients and applying denoising, outperforming traditional DPSGD methods.

Contribution

The authors propose gradient encoding and denoising techniques that improve privacy-utility trade-offs in differentially private deep learning models.

Findings

Achieves higher accuracy at the same privacy level compared to DPSGD.

Allows flexible choice of noise distributions for better privacy-utility balance.

Demonstrates significant privacy improvement on MNIST dataset.

Abstract

Deep learning models leak significant amounts of information about their training datasets. Previous work has investigated training models with differential privacy (DP) guarantees through adding DP noise to the gradients. However, such solutions (specifically, DPSGD), result in large degradations in the accuracy of the trained models. In this paper, we aim at training deep learning models with DP guarantees while preserving model accuracy much better than previous works. Our key technique is to encode gradients to map them to a smaller vector space, therefore enabling us to obtain DP guarantees for different noise distributions. This allows us to investigate and choose noise distributions that best preserve model accuracy for a target privacy budget. We also take advantage of the post-processing property of differential privacy by introducing the idea of denoising, which further improves the utility of the trained models without degrading their DP guarantees. We show that our mechanism outperforms the state-of-the-art DPSGD; for instance, for the same model accuracy of $96.1\%$ on MNIST, our technique results in a privacy bound of $\epsilon=3.2$ compared to $\epsilon=6$ of DPSGD, which is a significant improvement.