$\mathsf{OPA}$: One-shot Private Aggregation with Single Client Interaction and its Applications to Federated Learning

Harish Karthikeyan; Antigoni Polychroniadou

Paper 2024/723

: One-shot Private Aggregation with Single Client Interaction and its Applications to Federated Learning

Harish Karthikeyan, JPMorgan AI Research, JPMorgan AlgoCRYPT CoE

Antigoni Polychroniadou, JPMorgan AI Research, JPMorgan AlgoCRYPT CoE

Abstract

Our work aims to minimize interaction in secure computation due to the high cost and challenges associated with communication rounds, particularly in scenarios with many clients. In this work, we revisit the problem of secure aggregation in the single-server setting where a single evaluation server can securely aggregate client-held individual inputs. Our key contribution is the introduction of One-shot Private Aggregation () where clients speak only once (or even choose not to speak) per aggregation evaluation. Since each client communicates only once per aggregation, this simplifies managing dropouts and dynamic participation, contrasting with multi-round protocols and aligning with plaintext secure aggregation, where clients interact only once. We construct based on LWR, LWE, class groups, DCR and demonstrate applications to privacy-preserving Federated Learning (FL) where clients {speak once}. This is a sharp departure from prior multi-round FL protocols whose study was initiated by Bonawitz et al. (CCS, 2017). Moreover, unlike the YOSO (You Only Speak Once) model for general secure computation, eliminates complex committee selection protocols to achieve adaptive security. Beyond asymptotic improvements, is practical, outperforming state-of-the-art solutions. We benchmark logistic regression classifiers for two datasets, while also building an MLP classifier to train on MNIST, CIFAR-10, and CIFAR-100 datasets. We build two flavors of (1) from (threshold) key homomorphic PRF and (2) from seed homomorphic PRG and secret sharing. The threshold Key homomorphic PRF addresses shortcomings observed in previous works that relied on DDH and LWR in the work of Boneh et al. (CRYPTO, 2013), marking it as an independent contribution to our work. Moreover, we also present new threshold key homomorphic PRFs based on class groups or DCR or the LWR assumption.

Note: - Re-structured paper with simplified constructions, and additional experiments.

Metadata

Available format(s): PDF
Category: Applications
Publication info: Preprint.
Keywords: secure aggregation privacy-preserving federated learning non-interactive single round class groups
Contact author(s): harish @ nyu edu
antigonipoly @ gmail com
History: 2024-10-22: last of 2 revisions; 2024-05-11: received; See all versions
Short URL: https://ia.cr/2024/723
License: CC BY

BibTeX

@misc{cryptoeprint:2024/723,
      author = {Harish Karthikeyan and Antigoni Polychroniadou},
      title = {$\mathsf{{OPA}}$: One-shot Private Aggregation with Single Client Interaction and its Applications to Federated Learning},
      howpublished = {Cryptology {ePrint} Archive, Paper 2024/723},
      year = {2024},
      url = {https://eprint.iacr.org/2024/723}
}