Joint Coverless Steganography and Image Transformation for Covert Communication of Secret Messages

With the rapid development of internet applications, privacy protection of secret messages in covert communication has become increasingly important. To address the issue of attacks received in covert communication, such as eavesdropping attacks, steganalysis attacks, and tempering attacks, we propose a novel scheme that combines coverless steganography and image transformation for the covert communication of secret messages. Instead of using an image as the carrier to embed secret messages, our coverless steganography approach hides secret messages by exploiting a generative network to take it and a latent as input to synthesize a stego image, thus essentially avoiding both eavesdropping and typical steganalysis attacks. Furthermore, the generated pseudorealistic stego image with distortion vulnerabilities leads to tampering, which ultimately results in the receiver not being able to correctly extract secret information. In response to this problem, we design an image transformation method, which converts the stego image into another realistic image (i.e., camouflage image) and thus provides ambiguity for spoofing attackers while preventing stego images from being tampered with. Additionally, the authentication information is embedded into the camouflage image to obtain the transmitted authentication image (
$AuI$
) and can be used by the receiver to validate the completeness and authenticity of the camouflage image. Compared with related steganography methods in recent years, our proposed scheme synthesizes more high-quality stego images, and the stego image can be restored losslessly from the camouflage image. Most importantly, the PSNR of the camouflage image with authentication information (i.e.,
$AuI$
) is 31.78, which is enough to deceive the attacker; moreover, the secret message extraction rate can reach 100%.