Publications

Abstract : Universal Steganalysis can classify images without the knowledge of steganographic algorithms. This steganalysis will blindly classify an image as cover or not, but finding how much payload embedded, is still an open problem. This paper focuses on the above problem. Firstly, they use features from universal steganlysers and apply principal component analysis to improve the false positive rate. The above features are then used to estimate the payload by using support vector regression. The support vector machine classifier capable of assigning stego images to six popular steganographic algorithm after applying Principal Component Analysis: JP Hide amp; Seek, PVD, LSB flipping, Outguess, S-Tool and F5 is trained. This provides significantly more reliable results compared to their previous work on universal steganalysis. The performance is also evaluated by quantitative steganalysis for six steganographic algorithms.