Robust steganography resisting JPEG compression by improving selection of cover element

Abstract

Through minimizing embedding cost, modern adaptive steganography has gained unprecedented success in terms of its undetectability. However, due to Joint Photographic Experts Group (JPEG) compression, the secret bits hidden in the image transmitted over social network platform fail to be perfectly extracted, that remarkably limits its wide application in the real world. In this paper, by improving selection of cover element, we propose to design an enhanced robust steganographic algorithm to resist against JPEG compression. First, since that before and after JPEG compression the sign of Discrete Cosine Transform (DCT) coefficient is not easy to change, we devise cover element based on the sign of DCT coefficient. Furthermore, the selection of cover element is successfully improved with the help of post-processing operation analysis. Second, the embedding cost for each candidate DCT coefficient is calculated. Finally, dependent of both error correction algorithm and syndrome-trellis codes, a compression-resistant stego image is generated with minimum distortion. Numerical results verify that the robustness of our proposed steganographic algorithm is superior to that of current arts; meanwhile, the effectiveness of the algorithm is also verified over social network platform.