On the local convergence of GANs with differential Privacy: Gradient clipping and noise perturbation

Generative Adversarial Networks (GANs) are known to implicitly memorize details of sensitive data used to train them. To prevent privacy leakage, many approaches have been conducted. One of the most popular approaches is Differential Private Gradient Descent GANs (DPGD GANs), where the discriminator...

Full description

Saved in:
Bibliographic Details
Published in:Expert systems with applications 2023-08, Vol.224, p.120006, Article 120006
Main Authors: Azadmanesh, Maryam, Shahgholi Ghahfarokhi, Behrouz, Ashouri Talouki, Maede, Eliasi, Hussein
Format: Article
Language:English
Subjects:
Convergence
Differential Privacy
Generative Adversarial Network
Gradient Descent
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Generative Adversarial Networks (GANs) are known to implicitly memorize details of sensitive data used to train them. To prevent privacy leakage, many approaches have been conducted. One of the most popular approaches is Differential Private Gradient Descent GANs (DPGD GANs), where the discriminator’s gradients are clipped, and an appropriate random noise is added to the clipped gradients. In this article, a theoretical analysis of DPGD GAN convergence behavior is presented, and the effect of the clipping and noise perturbation operators on convergence properties is examined. It is proved that if the clipping bound is too small, it leads to instability in the training procedure. Then, assuming that the simultaneous/alternating gradient descent method is locally convergent to a fixed point and its operator is L-Lipschitz with L
ISSN:0957-4174
1873-6793
DOI:10.1016/j.eswa.2023.120006