Deep Neural Networks for Encrypted Inference with TFHE

Andrei Stoian; Jordan Frery; Roman Bredehoft; Luis Montero; Celia Kherfallah; Benoit Chevallier-Mames

Paper 2023/257

Deep Neural Networks for Encrypted Inference with TFHE

Andrei Stoian, Zama

Jordan Frery, Zama

Roman Bredehoft, Zama

Luis Montero, Zama

Celia Kherfallah, Zama

Benoit Chevallier-Mames, Zama

Abstract

Fully homomorphic encryption (FHE) is an encryption method that allows to perform computation on encrypted data, without decryption. FHE preserves the privacy of the users of online services that handle sensitive data, such as health data, biometrics, credit scores and other personal information. A common way to provide a valuable service on such data is through machine learning and, at this time, Neural Networks are the dominant machine learning model for unstructured data. In this work we show how to construct Deep Neural Networks (DNN) that are compatible with the constraints of TFHE, an FHE scheme that allows arbitrary depth computation circuits. We discuss the constraints and show the architecture of DNNs for two computer vision tasks. We benchmark the architectures using the Concrete stack, an open-source implementation of TFHE.

Metadata

Available format(s): PDF
Category: Applications
Publication info: Preprint.
Keywords: FHE Machine Learning Deep Learning TFHE Privacy
Contact author(s): andrei stoian @ zama ai
History: 2023-02-22: approved; 2023-02-22: received; See all versions
Short URL: https://ia.cr/2023/257
License: CC BY

BibTeX

@misc{cryptoeprint:2023/257,
      author = {Andrei Stoian and Jordan Frery and Roman Bredehoft and Luis Montero and Celia Kherfallah and Benoit Chevallier-Mames},
      title = {Deep Neural Networks for Encrypted Inference with {TFHE}},
      howpublished = {Cryptology {ePrint} Archive, Paper 2023/257},
      year = {2023},
      url = {https://eprint.iacr.org/2023/257}
}