Paper 2022/1674

Practical Multi-Key Homomorphic Encryption for More Flexible and Efficient Secure Federated Aggregation (preliminary work)

Alberto Pedrouzo-Ulloa, atlanTTic Research Center, Universidade de Vigo
Aymen Boudguiga, CEA List, Université Paris-Saclay
Olive Chakraborty, CEA List, Université Paris-Saclay
Renaud Sirdey, CEA List, Université Paris-Saclay
Oana Stan, CEA List, Université Paris-Saclay
Martin Zuber, CEA List, Université Paris-Saclay
Abstract

In this work, we introduce a lightweight communication-efficient multi-key approach suitable for the Federated Averaging rule. By combining secret-key RLWE-based HE, additive secret sharing and PRFs, we reduce approximately by a half the communication cost per party when compared to the usual public-key instantiations, while keeping practical homomorphic aggregation performances. Additionally, for LWE-based instantiations, our approach reduces the communication cost per party from quadratic to linear in terms of the lattice dimension.

Metadata
Available format(s)
PDF
Category
Applications
Publication info
Preprint.
Keywords
Federated Learning Secure Federated Aggregation Multi-Key Homomorphic Encryption
Contact author(s)
apedrouzo @ gts uvigo es
aymen boudguiga @ cea fr
olive chakraborty @ cea fr
renaud sirdey @ cea fr
oana stan @ cea fr
martin zuber @ cea fr
History
2022-12-02: approved
2022-12-01: received
See all versions
Short URL
https://ia.cr/2022/1674
License
Creative Commons Attribution
CC BY

BibTeX

@misc{cryptoeprint:2022/1674,
      author = {Alberto Pedrouzo-Ulloa and Aymen Boudguiga and Olive Chakraborty and Renaud Sirdey and Oana Stan and Martin Zuber},
      title = {Practical Multi-Key Homomorphic Encryption for More Flexible and Efficient Secure Federated Aggregation (preliminary work)},
      howpublished = {Cryptology {ePrint} Archive, Paper 2022/1674},
      year = {2022},
      url = {https://eprint.iacr.org/2022/1674}
}
Note: In order to protect the privacy of readers, eprint.iacr.org does not use cookies or embedded third party content.