Selection-Channel-Aware Scale Co-occurrence Matrix Feature to Detect Content-Adaptive JPEG Steganography

Abstract

Content-adaptive image steganography constrains the embedding changes to the complex regions difficult to model by using heuristic embedding distortion function and STCs. It can achieve the good anti-detection ability, however, its selection channel is public. In this paper, the utilization of selection channel of content-adaptive JPEG steganography for steganalysis is studied and a selection-channel-aware scale co-occurrence matrix feature based on 2D Gabor filter is proposed. Firstly, the embedding change of content-adaptive JPEG steganography is analyzed. Then, the selection-channel-aware scale co-occurrence matrix based on 2D Gabor filter is proposed and the detailed extraction process is shown. In the experiments, the comparisons with the other steganalysis features show that the proposed steganalysis feature can obtain the competitive detection performances.