Abstract: Secure aggregation is an information-theoretic mechanism for gradient aggregation in federated learning, to aggregate the local user gradients without revealing them in the clear. In this ...