Question

A healthcare analytics team wants to build a predictive model using patient records held across five hospitals. Regulatory constraints prohibit centralizing the raw patient data in a single repository, but the hospitals are willing to contribute to a shared model. A privacy engineer is asked to recommend an architecture that lets the model learn from all datasets while keeping each hospital's raw records local. Which approach BEST satisfies these requirements?

Accepted Answer

Federated learning, where model parameters are trained locally at each hospital and only aggregated updates are shared with a central coordinator

Answer

Extract, pseudonymize, and copy each hospital's records into a central data lake, then train the model on the combined dataset

Answer

Apply static data masking to each hospital's dataset and email the masked extracts to the analytics team for consolidation

Answer

Grant the analytics team read-only ABAC access to each hospital's production database to query records remotely