Steganalysis-SPAM-FILTER

This paper presents a method for detection of steganographic method in spatial domain usiing LSB matching using comparison of results between different classifiers. First, arguments are provided for modeling the differences between adjacent pixels using firstorder and second-order Markov chains. Subsets of sample transition probability matrices are then used as features for a steganalyzer implemented by Support Vector Machines, Fisher Discriminant Analysis, Linear Discriminant Analysis and Fisher Discrminant Analysis(L1 norm). A set of 6000 images from boss database is used form training and testing purposes. We compare the results of using different orders of markov chain and also the result of different classifiers using false positive rate and accuracy.