文献类型：期刊文献

英文题名：Asymmetric Image Encryption-Hiding Scheme Based on Reversible Neural Network

作者：Liu, Min[1]; Ye, Guodong[2]

机构：[1] School of Management, Guangdong Ocean University, Zhanjiang, 524088, China; [2] Faculty of Mathematics and Computer Science, Guangdong Ocean University, Zhanjiang, 524088, China

年份：2024

外文期刊名：SSRN

收录：EI(收录号：20240219597)

语种：英文

外文关键词：Deep learning - Image compression - Image enhancement - Network coding - Network security - Neural network models

外文摘要：In this paper, a novel asymmetric image encryption-hiding scheme (AiEhS) using reversible neural network (RNN) is presented, which mainly use the strong learning ability of deep learning for compression and hiding of the secret plain image (SPI), thereby enhancing the efficiency of the encryption algorithm and improving the hiding effectiveness. Firstly, AiEhS employs an auto-encoder to compress the SPI and designs a new encryption method to encrypt the compressed image to get a cipher image, reaching the first layer of encryption. Secondly, a random meaningful carrier image is selected, and the above cipher image is then embedded into a carrier image using an RNN to obtain a carrier image hiding secrets (CiHS), thus realizing the second layer of hiding. Moreover, the proposed scheme AiEhS produces a pseudo-random sequence using a hyperchaotic map and constructs a new key model (NKM) to introduce plaintext dependency. Then, the keys are designed to be distributed by RSA encryption algorithm, thus effectively improving its security. In particular, pixels in the cipher image are decomposed, combined and scrambled to obtain another scrambled image, and then the trained RNN model is utilized to embed this scrambled image into a carrier image, resulting in a CiHS. Compared with the traditional compressive sensing based image hiding methods. ? 2024, The Authors. All rights reserved.