A Note on “Secure Quantized Training for Deep Learning”

Marcel Keller; Ke Sun

Paper 2023/1219

A Note on “Secure Quantized Training for Deep Learning”

Marcel Keller

, CSIRO's Data61

Ke Sun

, CSIRO's Data61, Australian National University

Abstract

Keller and Sun (ICML'22) have found a gap in the accuracy between floating-point deep learning in cleartext and secure quantized deep learning using multi-party computation. We have discovered that this gap is caused by a bug in the implementation of max-pooling. In this note, we present updated figures to support this conclusion. We also add figures for another network on CIFAR-10.

Metadata

Available format(s): PDF
Category: Implementation
Publication info: Preprint.
Keywords: Privacy-preserving machine learning secure multi-party computation
Contact author(s): mks keller @ gmail com
ke sun @ data61 csiro au
History: 2023-08-11: approved; 2023-08-11: received; See all versions
Short URL: https://ia.cr/2023/1219
License: CC BY

BibTeX

@misc{cryptoeprint:2023/1219,
      author = {Marcel Keller and Ke Sun},
      title = {A Note on “Secure Quantized Training for Deep Learning”},
      howpublished = {Cryptology ePrint Archive, Paper 2023/1219},
      year = {2023},
      note = {\url{https://eprint.iacr.org/2023/1219}},
      url = {https://eprint.iacr.org/2023/1219}
}