Image Steganalysis Algorithm Based on Improved Discernibility Matrix

Aimed to improve the detection efficiency of steganalysis system, the present study proposes an ISAIDM-L image steganalysis algorithm that is based on improved discernibility matrix, according to the high dimensions, and the intensive occupation of discernibility matrix on the contiguous memory. First, to solve the problem of the intensive occupation on contiguous memory, the consistent table convertion ways have been analyzed before the construction ways of discernibility matrix have been proposed which is based on improved consistent decision table; second, to reduce the dimension of statistical features and computational complex with relevance between statistical features removed, an image steganalysis algorithm has been devised which based on the improved discernibility matrix; third, while using the proposed algorithm to reduce the statistical features extraced from different image databases, image formats and different domains, SVM classifier takes classification training tests on steganographic tools such as Cox, Piva, jphide and MB2, etc. Experimental results indicate that the proposed steganalysis algorithm has better universality, stability and effectiveness while it gets significant improvement in time efficiency, feature accuracy, etc.