Multiclass Detector of Current Steganographic Methods for JPEG Format

摘要: The aim of this paper is to construct a practical forensic steganalysis tool for JPEG images that can properly analyze single- and double-compressed stego classify them selected current steganographic methods. Although some the individual modules steganalyzer were previously published by authors, they never tested as complete system. fusion brings its own challenges problems whose analysis solution one goals paper. By determining stego-algorithm, provides first step needed extracting secret message. Given image, detector assigns it six popular algorithms. detection based on feature extraction supervised training two banks multiclassifiers realized using support vector machines. For accurate classification single-compressed images, separate multiclassifier trained each quality factor from certain range. Another bank same range primary factors. image under investigation analyzed preclassifier detects cases double compression estimates quantization table. It then sends appropriate or double-compression multiclassifier. error estimated more than 2.6 million images. also unseen methods examine ability generalize.