https://github.com/illidanlab/dpgan

Source code of paper "Differentially Private Generative Adversarial Network"
https://github.com/illidanlab/dpgan

Last synced: 3 months ago
JSON representation

Source code of paper "Differentially Private Generative Adversarial Network"

• Host: GitHub
• URL: https://github.com/illidanlab/dpgan
• Owner: illidanlab
• Created: 2017-08-30T15:10:49.000Z (about 7 years ago)
• Default Branch: master
• Last Pushed: 2018-11-29T19:29:33.000Z (almost 6 years ago)
• Last Synced: 2024-04-08T02:27:02.520Z (7 months ago)
• Language: Python
• Homepage:
• Size: 21.9 MB
• Stars: 66
• Watchers: 6
• Forks: 28
• Open Issues: 3
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• awesome-data-synthesis - DP-GAN 2 - Source code of paper "Differentially Private Generative Adversarial Network" - [Paper](https://arxiv.org/abs/1802.06739) (Data-driven methods / Tabular)

README

        
# Differentially Private Generative Adversarial Network

Liyang Xie, Kaixiang Lin, Shu Wang, Fei Wang, Jiayu Zhou

## Paper link

https://arxiv.org/abs/1802.06739

## Abstract

Generative Adversarial Network (GAN) and its variants have re-

cently attracted intensive research interests due to their elegant

theoretical foundation and excellent empirical performance as gen-

erative models. These tools provide a promising direction in the

studies where data availability is limited. One common issue in

GANs is that the density of the learned generative distribution

could concentrate on the training data points, meaning that they

can easily remember training samples due to the high model com-

plexity of deep networks. This becomes a major concern when

GANs are applied to private or sensitive data such as patient medi-

cal records, and the concentration of distribution may divulge criti-

cal patient information. To address this issue, in this paper we pro-

pose a differentially private GAN (DPGAN) model, in which we

achieve differential privacy in GANs by adding carefully designed

noise to gradients during the learning procedure. We provide rig-

orous proof for the privacy guarantee, as well as comprehensive

empirical evidence to support our analysis, where we demonstrate

that our method can generate high quality data points at a reason-

able privacy level.