International Journal “Information Theories and Applications”, Vol. 24, Number 3, © 2017 55 MULTICLASS DETECTOR FOR MODERN STEGANOGRAPHIC METHODS Dmytro Progonov Abstract: Creation of advanced steganalysis methods for reliably detection of hidden messages in widespread multimedia files, such as digital images, is topical task today. One of the key requirements to such methods is ability to reveal the stego files even in case of limited or absent information relating to used embedding methods. For solving this task there was proposed the multiclass stegdetector, based on applying the powerful methods of digital image structural analysis. Obtained earlier results confirmed the high efficiency of proposed stegdetector by message hiding in cover image’s transformation domain. There is conducted analysis of stegdetector performance in case of message hiding according to advanced adaptive steganogaphic methods, such as HILL, MiPOD and Synch algorithms. It is shown that usage the “extended” cover image model, includes not only statistical, but also correlation and fractal features, gives opportunity to improve the detection accuracy of stegdetector in most difficult cases of image steganalysis. Keywords: digital images steganalysis, adaptive embedding methods, multiclass stegdetector. ITHEA Keywords: K.6.5 Management of computing and information systems. Security and Protection; I.4.10 Image processing and computer vision. Image representation. Introduction Protection of private as well as state-owned sensitive information is urgent problem today. Considerable quantity of freely available malware, ransomware and operation system’s backdoors packets allow any users of Internet to create the personal toolbox for attacking not only private computers, but also the information infrastructures systems of governmental agencies as well as private corporations. Distinctive feature of such attacks is wide usage of complicated methods for creation the hidden communication [Cisco, 2015; Cisco, 2016; FireEye, 2015]. These channels are integrated into information flows in telecommunication systems, like email, social networks, file sharing networks, which complicates the issue of theirs detection and counteraction by state security analytics agencies. It is worth noting that in most cases information relating to data embedding process is limited or even absent. Therefore, applying of known signature or statistical steganalysis methods does allow providing the high accuracy of stego files detection. That is why development of new steganalysis approaches,