Secure Aggregation

This project enhances the CAFEIN platform by integrating secure aggregation capabilities into federated learning (FL) and federated analytics. Secure aggregation is a cryptographic technique that enables multiple participants to combine their data contributions—whether model updates or statistical metrics—while ensuring that only the final aggregated result is revealed, not individual inputs.

This approach:

• protects privacy by preventing exposure of individual contributions;
• mitigates membership inference attacks that aim to extract information about individual participants.

In FL, secure aggregation ensures that individual model updates remain private while still enabling collaborative training. Federated analytics allows the computation of statistical metrics (e.g., histograms, counts, averages) across multiple data sources without exposing raw data.

Our implementation for CAFEIN is designed to balance security and privacy guarantees, computational efficiency, and scalability, making secure aggregation practical for real-world healtcare federated applications.