53. Privacy-preserving deep learning

Protect sensitive training and inference data with minimization, de-identification, differential privacy, secure computation, and federated learning. Evaluate privacy leakage and the utility cost of protective measures.