Dropout-tolerant secure aggregation for federated model updates

What changed

Combined pairwise pseudorandom masks, threshold secret sharing and double masking to recover sums of high-dimensional client updates despite dropouts while limiting disclosure of individual updates.

Scientific scope

Combined pairwise pseudorandom masks, threshold secret sharing and double masking to recover sums of high-dimensional client updates despite dropouts while limiting disclosure of individual updates. Pinned 2016 workshop protocol, not an independently checked full 2017 CCS security proof. Different server threat models require different thresholds; practical unauthenticated bootstrap assumes non-malicious key exchange and remains vulnerable to active MITM. Arbitrarily malicious colluding users/server deferred to full version. Aggregate output still reveals aggregate information and is not DP; dropout tolerance is not blanket participant-security.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Pinned 2016 workshop protocol, not an independently checked full 2017 CCS security proof. Different server threat models require different thresholds; practical unauthenticated bootstrap assumes non-malicious key exchange and remains vulnerable to active MITM. Arbitrarily malicious colluding users/server deferred to full version. Aggregate output still reveals aggregate information and is not DP; dropout tolerance is not blanket participant-security.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• Practical Secure Aggregation for Federated Learning on User-Held Data

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction