A novel steganalysis of LSB matching based on kernel FDA in grayscale images

To detect presence of LSB matching blindly, a novel steganalysis is proposed. First, image segmentation is employed to separate image into different domains. Second, statistic property of node degree for minimum spanning tree (MST) in random domain is analyzed. And third, local image complexity is proposed to describe concrete domain situation, and image features are also extracted accordingly. Simulation results demonstrate that the proposed algorithm can achieve higher detection probability than existent ones on both uncompressed and compressed image formats, especially low embedding rate.