Constructing local information feature for spatial image steganalysis

Feature is a key part for steganalysis. In this paper we propose a spatial feature set for image steganalysis, named Local Information Feature (LIF), to increase the diversity of spatial steganalysis feature and improve its performance. It also provide a heuristic framework for designing steganalysis feature through 3 steps. It first collects local information from its local region consisting of adjacent pixels. Then according to certain rules, it maps each pixel to its corresponding local type by its local information. Finally, the feature set is formed by adaptive weighted statistical histograms of local types. We design two schemes for LIF, each of which can generate different feature sets using different methods of local information computing. Experimental results show that our feature is effective for detecting stego images embedded by adaptive steganography. We also discussed some possible method to extent the feature designing based on LIF.