Robust Quantity-Aware Aggregation for Federated Learning. (arXiv:2205.10848v2 [cs.CR] UPDATED)

Federated learning (FL) enables multiple clients to collaboratively train
models without sharing their local data, and becomes an important
privacy-preserving machine learning framework. However, classical FL faces
serious security and robustness problem, e.g., malicious clients can poison
model updates and at the same time claim large quantities to amplify the impact
of their model updates in the model aggregation. Existing defense methods for
FL, while all handling malicious model updates, either treat all quantities
benign or simply ignore/truncate the quantities …

aggregation amplify aware claim clients data federated learning framework impact important large local machine machine learning malicious privacy problem robustness security serious serious security sharing train updates