ISACA AI Fundamentals Practice Test 2026 – All-in-One Resource to Master Your Certification!

Federated learning trains the model across many devices while keeping the actual data on each device. Each device uses its local data to update a local copy of the model, and only these updates (not the raw data) are sent to a central server. The server aggregates these updates to form a new global model and then distributes it back to the devices. This approach preserves user data on the device, addressing privacy while still enabling a shared, improving model as more devices participate.

Data minimization focuses on limiting what data is collected, not on coordinating a global model. Encryption protects data in transit or at rest but doesn’t by itself enable training with data remaining on devices. Differential privacy adds noise to protect individual contributions, but it isn’t defined by keeping data locally while producing a global model.