Improving the Embedding Strategy for Batch Adaptive Steganography

Abstract

Recent works have demonstrated that images with more texture regions should be selected as the sub-batch of covers to carry the total message when applying batch steganography to adaptive steganography and the core challenge of which is how to evaluate the texture complexity of image accurately according to the need of steganography security. In this paper, we first propose three methods for measuring the texture complexity of image to select images with highly textured content, then put forward our universal embedding strategy for batch adaptive steganography in both spatial and JPEG domain. To assess the security of embedding strategies for batch adaptive steganography, we use a pooling steganalysis method based majority decision for the omniscient Warden, who informed by the average payload, embedding algorithm and cover source. Given a batch of images, our proposed embedding strategy is to select images with largest residual values to carry the total message, which is named max-residual-greedy (MRG) strategy. Experimental results show that the proposed embedding strategy outperforms the previous ones for batch adaptive steganography.