A Steganalysis Based on OC-SVM and Two Clustering Methods
Hai Bin Yang, Dun Tao Guo, Peng Cheng Luo · Applied Mechanics and Materials · 2012
In this paper, a novel and efficient steganalysis based on improved K-means clustering using One-Class Support Vector Machine (OC-SVM) was proposed to blindly determine the existence of hidden messages in an image. The performance of sample clustering is concerned in the OC-SVM with multi-sphere. In previous work, the K-means was mainly used to create such multi-sphere by clustering. But the traditional K-means depends on initial clustering centers and ended local minimum value. So, to improve the clustering performance of K-means, the SA is employed into the scheme, which can create more reasonable multi-sphere by finding global optimum solutions in the clustering process. Simulation results with the chosen feature set and well-known steganographic techniques indicated that our approach was able to afford reasonable accuracy to distinguish between covers and stego images.