Assess ability to design a privacy-preserving federated learning system that integrates differential privacy and secure aggregation. Test selection and tuning of DP mechanisms (local vs. central, Gaussian/Laplace), privacy accounting and composition (ε, δ), secure aggregation choices (MPC, homomorphic encryption), client sampling and dropout handling, communication/compute trade-offs, threat models, robustness to attacks, and evaluation of utility-versus-privacy in deployment scenarios.