Entropy Feature Based on 2D Gabor Wavelets for JPEG Steganalysis

Abstract

To improve the detection accuracy for adaptive JPEG steganography which constrains embedding changes to image texture regions difficult to model, a new steganalysis feature based on the Shannon entropy of 2-dimensional (2D) Gabor wavelets is proposed. For the proposed feature extraction method, the 2D Gabor wavelets which have certain optimal joint localization properties in spatial domain and in the spatial frequency are employed to capture the image texture characteristics, and then the Shannon entropy values of image filtering coefficients are used as steganalysis feature. First, the decompressed JPEG image is filtered by 2D Gabor wavelets with different scale and orientation parameters. Second, the entropy features are extracted from all the filtered images and then they are merged according to symmetry. Last, the ensemble classifier trained by entropy features is used as the final steganalyzer. The experimental results show that the proposed feature can achieve a competitive performance by comparing with the state-of-the-art steganalysis features for the latest adaptive JPEG steganography algorithms.