Feature selection for steganalysis based on modified Stochastic Diffusion Search using Fisher score

The steganalysis process comprises of two major components feature extraction and classification based on the extracted features. The high dimension of feature sets used for steganalysis makes classification a complex and time-consuming process. A novel feature selection algorithm (SDSFS) for steganalysis is proposed in this paper. SDSFS is a filter-type feature selection algorithm which selects reduced feature set based on Stochastic Diffusion Search. The Stochastic Diffusion Search, a generic population-based search method, has been modified and adopted successfully for steganalytic feature selection in this work. The algorithm uses separability of feature vectors as hypothesis and Fisher score is used as separability measure. To show the usefulness and effectiveness of SDSFS, experiments were conducted on two different feature sets used for steganalysis. The experimental results show that the proposed feature selection effectively reduces the dimensionality of the features and improves the detection accuracy of the steganalysis process.