A New Design Method about the Universal Steganalysis Classifier in Digital Image
Yanping Tan, Yong Luo, Wenhua He, Yan Xiong, Xianxiang Chen · 2020
Based on a variety of commonly used steganographic algorithms in the spatial domain of digital images, a new design method is proposed for the universal steganalysis classifier. The main idea is that we use steganalysis to extract features from the cover image set and their corresponding steganographic image set for a certain kind of steganographic algorithm, and those features are used for training the corresponding classifier. Then the classifier is used to test the remaining steganographic algorithms respectively and until the cross test is completed successively. According to the cross test results, we chose three adaptive steganographic algorithms from various steganographic algorithms. Then we combine features of the selected three steganographic algorithms to train a universal steganalysis classifier in digital image. When the method used by the image to embed secret information is unknown, we can give priority to use the universal steganalysis classifier to determine whether they contain secret messages, thereby improving the overall work efficiency. Experimental results demonstrate the feasibility of our proposed method for the universal steganalysis classifier in digital image.