Breaking Secure Aggregation: Label Leakage from Aggregated Gradients in Federated Learning

Open in new window