← Back to glossary
🧮

Secure Aggregation

Privacy

Encryption Differential Privacy Federated Learning Secure Multi‑Party Computation

Cryptographic protocol allowing computation of aggregates without revealing individual contributions.

Definition

Secure aggregation hides individual client updates (or values) while enabling the server to compute a sum/average. It’s often used with federated learning and can be implemented using secret sharing and encryption.

In plain English Cryptographic protocol allowing computation of aggregates without revealing individual contributions.

Why this matters

Why it matters: It reduces data exposure during analytics or training, lowering re-identification risk.

Example

Example: Clients encrypt updates; the server can only decrypt the aggregate once enough participants contribute.