IntroStep 1 of 3

Bonawitz et al. Secure Aggregation

Practical Secure Aggregation for Privacy-Preserving Machine Learning

arXiv:1611.04482

This interactive demonstration walks you through the complete Bonawitz secure aggregation protocol, exactly as described in the original paper. You'll see how multiple clients can collaboratively compute an aggregate of their private data without revealing individual values.

Privacy Guaranteed

Server learns only the aggregate sum, never individual inputs

Dropout Resilient

Protocol succeeds even if some clients disconnect

Cryptographically Secure

Uses Diffie-Hellman key exchange and Shamir Secret Sharing