Privacy-Preserving Deep Neural Networks Using Pixel-Based Image Encryption Without Common Security Keys

Abstract

We present a novel privacy-preserving scheme for deep neural networks (DNNs) that enables us not to only apply images without visual information to DNNs but to also consider the use of independent encryption keys, for both training and testing images for the first time. In this paper, a novel pixel-based image encryption method, which considers maintaining the properties of original images, is first proposed for privacy-preserving DNNs. For training, a DNN model is trained with images encrypted by using the proposed method under the use of independent keys. For testing, the model enables us to applied both encrypted images and plain images for image classification. Therefore, there is no need to manage the keys. In an experiment, the proposed method is applied to a well-known network, deep residual networks, for image classification. The experimental results demonstrate that the proposed method with independent encryption keys has robustness against ciphertext-only attack (COA) and can provide almost the same classification performance as that of using plain images. Moreover, the results confirm that the proposed scheme is able to classify plain images as well as encrypted images.