In the past few years, the Generative Adversarial Network (GAN), which proposed in 2014, has achieved great success. There have been increasing research achievements based on GAN in the field of computer vision and natural language processing. Image steganography is an information security technique aiming at hiding secret messages in common digital images for covert communication. Recently, research on image steganography has demonstrated great potential by introducing GAN and other neural network techniques. In this paper, we review the art of steganography with GANs according to the different strategies in data hiding, which are cover modification, cover selection, and cover synthesis. We discuss the characteristics of the three strategies of GAN-based steganography and analyze their evaluation metrics. Finally, some existing problems of image steganography with GAN are summarized and discussed. Potential future research topics are also forecasted.

Schemes of reversible data hiding in encrypted domain (RDH-ED) based on symmetric or public key encryption are mainly applied in the scenarios of end-to-end communication. To provide security guarantees for the multi-party scenarios, a RDH-ED scheme for secret image sharing based on Chinese remainder theorem (CRT) is presented. In the application of ( $t$ , $n$ ) secret image sharing, an image is first shared into $n$ different shares of ciphertext. Only when not less than $t$ shares obtained, can the image be reconstructed. In our scheme, additional data could be embedded into the image shares. To realize data extraction from the image shares and the reconstructed image separably, two data hiding methods are proposed: one is homomorphic difference expansion in encrypted domain (HDE-ED) that supports data extraction from the reconstructed image by utilizing the addition homomorphism of CRT secret sharing; the other is difference expansion in image shares (DE-IS) that supports the data extraction from the marked shares before image reconstruction. Experimental results demonstrate that the proposed scheme could not only maintain the security and the threshold function of secret sharing system, but also obtain a better reversibility and efficiency compared with most existing RDH-ED algorithms. The maximum embedding rate of HDE-ED could reach 0.500 bits per pixel and the average embedding rate of DE-IS could reach 0.4652 bits per pixel.

In this paper, a novel data-driven information hiding scheme called “generative steganography by sampling” (GSS) is proposed. Unlike in traditional modification-based steganography, in our method the stego image is directly sampled by a powerful generator: no explicit cover is used. Both parties share a secret key used for message embedding and extraction. The Jensen-Shannon divergence is introduced as a new criterion for evaluating the security of generative steganography. Based on these principles, we propose a simple practical generative steganography method that uses semantic image inpainting. The message is written in advance to an uncorrupted region that needs to be retained in the corrupted image. Then, the corrupted image with the secret message is fed into a Generator trained by a generative adversarial network (GAN) for semantic completion. Message loss and prior loss terms are proposed for penalizing message extraction error and unrealistic stego image. In our design, we first train a generator whose training target is the generation of new data samples from the same distribution as that of existing training data. Next, for the trained generator, backpropagation to the message and prior loss are introduced to optimize the coding of the input noise data for the generator. The presented experiments demonstrate the potential of the proposed framework based on both qualitative and quantitative evaluations of the generated stego images.

Steganography is a technique for publicly transmitting secret information through a cover. Most of the existing steganography algorithms are based on modifying the cover image, generating a stego image that is very similar to the cover image but has different pixel values, or establishing a mapping relationship between the stego image and the secret message. Attackers will discover the existence of secret communications from these modifications or differences. In order to solve this problem, we propose a steganography algorithm ISTNet based on image style transfer, which can convert a cover image into another stego image with a completely different style. We have improved the decoder so that the secret image features can be fused with style features in a variety of sizes to improve the accuracy of secret image extraction. The algorithm has the functions of image steganography and image style transfer at the same time, and the images it generates are both stego images and stylized images. Attackers will pay more attention to the style transfer side of the algorithm, but it is difficult to find the steganography side. Experiments show that our algorithm effectively increases the steganography capacity from 0.06 bpp to 8 bpp, and the generated stylized images are not significantly different from the stylized images on the Internet.

Deep steganography utilizes the powerful capabilities of deep neural networks to embed and extract messages, but its reliance on an additional message extractor limits its practical use due to the added suspicion it can raise from steganalyzers. To address this problem, we propose StegaINR, which utilizes Implicit Neural Representation (INR) to implement steganography. StegaINR embeds a secret function into a stego function, which serves as both the message extractor and the stego medium for secure transmission on a public channel. Recipients only need to use a shared key to recover the secret function from the stego function, allowing them to obtain the secret message. Our approach employs continuous functions, enabling it to handle various types of messages. To our knowledge, this is the first work to introduce INR into steganography. We perform evaluations on image, climate data, and Neural Radiance Field (NeRF) synthetic dataset to test our method in different deployment contexts.