FedNova

Federated learning accuracy can hide critical model failures, study finds

A research project comparing federated learning algorithms for network intrusion detection revealed that high global accuracy can mask poor performance on minority data silos. The study found that FedAvg achieved 96% gl…