Title:	Privacy Preserving Federated Learning
DNr:	Berzelius-2024-103
Project Type:	LiU Berzelius
Principal Investigator:	Sargam Gupta <sargam.gupta@umu.se>
Affiliation:	Umeå universitet
Duration:	2024-04-01 – 2024-10-01
Classification:	10201
Keywords:

Abstract

I plan to work on a privacy-preserving framework for the Deep Leakage from Gradients attack which is quite a popular attack in Federated Learning literature. Currently, I am specifically looking at the mitigation for this particular attack on the " DENSE: Data-Free One-Shot Federated Learning" paper. I also plan to experiment more with the federated settings and different privacy models like k-anonymity and differential privacy.