Define protected units and threats, bound contributions, review release accounting and test data-lineage controls.
Learning path
- Privacy-aware ML: threat model, data minimization and purpose
- Privacy units and bounded contribution: count people, not events
- Private aggregate releases: sensitivity, budget and query control
- Federated training: local data, update leakage and aggregation boundaries
- Model leakage evaluation: membership risk, memorization and limits
- Privacy operations: deletion lineage, access gates and release review
- Project: design a bounded learner-activity release
Connected foundations
Use the earlier data and model boundaries as prerequisites.
Further paths
Practice
Continue into another subject: Synthetic Data & Simulation Tutorial.