Assess students' ability to apply differential privacy mechanisms in federated learning to achieve a practical balance between user privacy and model utility; tasks include choosing appropriate DP approaches (local vs. central, DP-SGD), calibrating privacy budgets (ε, δ) and noise/clipping parameters, designing secure aggregation and communication strategies for heterogeneous clients, empirically and theoretically evaluating privacy-utility trade-offs on model metrics, and justifying design decisions under real-world constraints and threat models.