Steganalysis of homogeneous-representation based steganography for high dynamic range images

Abstract

The Homogeneous-Representation Based Steganography (HRBS) embeds secret information into High Dynamic Range (HDR) images by encoding different homogeneous representations of pixels. After embedding information, the original cover HDR image can be completely recovered by the receiver. However, there is no corresponding steganalytic algorithm for such steganography. In the context of this, we propose a steganography detection algorithm based on non-zero homogeneity index, which is generally applicable to homogeneous-representation based steganography, and three specific stego pixel ratio estimation algorithms based on the least squares method, which are applicable to three typical HRBS algorithms respectively. The homogeneity indexes of pixels in the cover HDR image are usually zero, while the homogeneity indexes of pixels in the stego HDR image may be changed due to information embedding. Therefore, the proposed steganography detection algorithm determines the existence of secret information by detecting whether there are pixels with non-zero homogeneity index in the investigated image. Then, according to the relationship between the stego pixel ratio and the number of pixels with abnormal homogeneity index in the stego image, the proposed stego pixel ratio estimation algorithms use the least squares method to estimate the stego pixel ratio of stegos embedded by three typical HRBS algorithms. The experimental results show that when the length of the embedded secret message reaches 7 bits, the proposed steganography detection algorithm can correctly detect the stego image with a probability of over 99%. And the proposed stego pixel ratio estimation algorithms can achieve quite good performance when targeting against three typical HRBS algorithms.