Combining Homomorphic Encryption and Differential Privacy in Federated Learning for Model Inspection and Availability
A federated-learning design combines homomorphic encryption and differential privacy, beating DP-only utility and privacy on FEMNIST.