Privacy-Preserving Image Classification Using Isotropic Network

TL;DR

This paper introduces a privacy-preserving image classification approach utilizing encrypted images and isotropic networks like vision transformers, achieving high accuracy without needing adaptation networks and supporting image compressibility.

Contribution

It presents a novel method combining encrypted images with isotropic networks for privacy-preserving classification, eliminating the need for adaptation networks and handling compressible images.

Findings

Outperforms state-of-the-art methods in accuracy and robustness.

Works effectively with encrypted and compressible images.

Applicable to vision transformer and ConvMixer networks.

Abstract

In this paper, we propose a privacy-preserving image classification method that uses encrypted images and an isotropic network such as the vision transformer. The proposed method allows us not only to apply images without visual information to deep neural networks (DNNs) for both training and testing but also to maintain a high classification accuracy. In addition, compressible encrypted images, called encryption-then-compression (EtC) images, can be used for both training and testing without any adaptation network. Previously, to classify EtC images, an adaptation network was required before a classification network, so methods with an adaptation network have been only tested on small images. To the best of our knowledge, previous privacy-preserving image classification methods have never considered image compressibility and patch embedding-based isotropic networks. In an experiment, the proposed privacy-preserving image classification was demonstrated to outperform state-of-the-art methods even when EtC images were used in terms of classification accuracy and robustness against various attacks under the use of two isotropic networks: vision transformer and ConvMixer.