A Universal Steganalysis Method for JPEG Images Based on Semi-supervised Learning
Shangping Zhong · Computer Technology and Development · 2009
At present,the universal steganalysis methods for JPEG images are based on supervised learning,the key technologies of these methods inclue image feature extraction and classifier design.Proposes a novel classifier which based on semi-supervised learning EM algorithm,the classifier makes use of a great quantity of unlabeled samples combined with a small quantity of labeled samples.Aimed at the popular JPEG steganography technologies: Outguess and F5,compared with this paper's method and the supervised learning method,the experimental results show that the proposed method can obtain good performance when a large number of labeled samples can't be obtained.Therefore,our method can improve the practicability for JPEG image universal steganalysis.