FedDyna: Privacy-Preserving Federated Learning with Dynamic Noise Adaptation and Structural Bias Alignment for Non-IID Environments
FedDyna is proposed, a novel framework that uses an adaptive noise injection mechanism to enhance privacy in non-IID data environments and effectively demonstrates the significant divergence between dynamic and static noise under non-IID settings.
Guijuan Wang, Zhiyu Zuo, An-Ming Dong et al.
· 0 citations