Leak and Learn: An Attacker's Cookbook to Train Using Leaked Data from Federated Learning

no code implementations • 26 Mar 2024 • Joshua C. Zhao, Ahaan Dabholkar, Atul Sharma, Saurabh Bagchi

We demonstrate the effectiveness of both GI and LLL attacks in maliciously training models using the leaked data more accurately than a benign federated learning strategy.

Federated Learning