Aiming at the problem of the weak avalanche effect in the recently proposed deep learning image encryption algorithm, this paper analyzes the causes of weak avalanche effect in the neural network of Cycle-GAN step-by-step process and proposes an image encryption algorithm combining the traditional diffusion algorithm and deep learning neural network. In this paper, first, the neural network is used for image scrambling and slight diffusion, and then the traditional diffusion algorithm is used to further diffuse the pixels. The experiment in satellite images shows that our algorithm, with the help of the further diffusion mechanism, can compensate for the weak avalanche effect of Cycle-GAN-based image encryption and can change a pixel value to the original image, and the number of pixel change rate (NPCR) and unified average changing intensity (UACI) values can achieve 99.64% and 33.49%, respectively. In addition, our method can effectively encrypt the image where the encrypted image with high information entropy and low pixel correlation is obtained. The experiment on data loss and noise attack declares our method can identify the types and intensity of attacks. What is more, the key space is big enough, and the key sensitivity is high while the key has a certain randomness.

In the digital era, sharing pictures on social media has become a common privacy issue. To prevent private images from being eavesdropped on and destroyed, developing secure and efficient image steganography, image cryptography, and image authentication has been difficult. Deep learning provides a solution for digital image security. First, we make an overall conclusion on deep learning applications in image steganography to generate five aspects: the cover image, stego-image, embedding change probabilities, coverless steganography, and steganalysis. Second, we also combine and compare deep learning methods used in six aspects: image cryptography from image compression, image resolution improvement, image object detection and classification, key generation, end-to-end image encryption, and image cryptoanalysis. Third, we collect deep learning methods in image authentication from five perspectives: image forgery detection, watermarked image generation, image watermark extraction and detection, image watermarking attack, and image watermark removal. Finally, we summarize future research directions of deep learning utilization in image steganography, image cryptography, and image authentication.

No abstract

No abstract

To realize image information hiding, we train a multi-layers autoencoder includes an encoder and a decoder. Using encoder architecture, we input the secret color image and generate the cover image where the size stays the same directly. In addition, we reconstruct the secret color image by inputting the stego images passing on the decoder which accomplish the secret image blind extraction. Our method realizes a high resolution and high payload color image steganography. Experimental results show that we can realize the high capacity blind specified secret images information hiding automatically by our method.