Multi-classification model for image steganalysis
Vladimir Banoci, Gabriel Bugár, Martin Broda, Dušan Levický · International Symposium ELMAR · 2013
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.