Multi-classification model for image steganalysis

This paper presents results of multi-classification and cross-validation of tested steganalysis method in static images in JPEG format. Steganalysis methods are used for revealing a secret communication conducted by different steganographic tools. The steganalysis algorithm analyzes changes in statistical parameters of the images using Feature Based Steganalysis. The multi-classification is ability of proposed system to identify an applied steganography methods and cross-validation is defined as steganalytic model´s detection efficiency of steganography methods that were not used in training phase of the model. Testing was performed for different length of statistical features´ vector and for different size of embedded secret message. The results are also contributing in design of blind steganography system to detect a new steganography tools.