Resisting Backdoor Attacks in Federated Learning via Bidirectional Elections and Individual Perspective
This work proposes a backdoor defense for FL from individual perspectives, which excludes infected model updates while maintaining main task accuracy.
Feb 28, 2024
