Framework for improving the security performance of ordinary distortion functions of JPEG steganography

Abstract

Digital steganography is a new approach for secure communication. Via using it, the sender and the receiver can easily exchange secret message on the Internet without arousing any suspicion. Previously, a lot of ordinary distortion functions for joint photographic experts group (JPEG) steganography have been presented, which can guide the message embedding in the non-zero alternating current (AC) discrete cosine transform (DCT) coefficients of JPEG image. In this paper, we present a framework for improving the security performance of these distortion functions. In our new framework, these ordinary distortion functions are not restricted to evaluating the distortion values of non-zero AC DCT coefficients any more, and their coverage areas will be extended to all DCT coefficients, including the direct current (DC) coefficients and all the zero and non-zero AC coefficients. All the coefficients in JPEG image are divided into two groups: changeable group (CG) and reserve group (RG), respectively. The coefficients that may result in less detectable distortion are grouped into CG and the rest into RG. Via associating the distortion values to coefficients in CG and RG with different strategies, a series of new advanced distortion functions can be generated. The experimental results demonstrate that while applying these advanced distortion functions to JPEG steganography, the statistical characteristics of the carrier image will be preserved better than the prior art, and consequently secure JPEG steganographic schemes can easily be obtained.