Byzantine-Robust Federated Learning Based on Multi-center Secure Clustering
Xiaosong Liu, Chungen Xu, Pan Zhang · 2023
In the backdrop of growing social apprehensions regarding data privacy and security, Federated Learning (FL) stands out as a cornerstone in modern machine learning. However, FL grapples with challenges stemming from data heterogeneity and potential security breaches. While techniques such as FedSEM enhance multi-center aggregation by employing Expectation Maximization (EM) methods to match clients with centers, leading to improved performance in non-IID environments, there remains a critical gap in ensuring optimal privacy and security measures.