The recovery scheme of computer-generated holography encryption–hiding images based on deep learning

Abstract

In order to improve the security of ciphertext images in transmission, a scheme of using neural network is proposed to restore the encryption–hiding images based on chaotic iris phase mask and computer-generated holography (CGH), which solve the great difficulty of illegal attacks in the symmetric–asymmetric hybrid encryption system. Firstly, the ciphertext image based on CGH and chaotic iris phase mask is generated, and the ciphertext image is hidden into a carrier image. A large number of hidden image and plaintext image pairs are produced as a dataset. Next, the neural network is built by continuous training and testing. Finally, the established neural network can fit the mapping relationship between the hidden image and the plaintext image. It is not necessary to extract the ciphertext image from the hidden image before decrypting. We compare the image recovered by the neural network with the plaintext image. The experimental results show that the average cross-correlation coefficient (CC) is 0.994, the average peak signal-to-noise ratio (PSNR) is 70.2 dB, and the average structural similarity (SSIM) is 0.929. This scheme can directly realize the decryption of the hidden image. The scheme (encryption–decryption–hiding) are elaborated in detail. The simulation experiments show that the scheme is feasible and has good robustness.