Blind image steganalysis based on texture and noise features

Hidden information embedded in digital images will inevitably lead to changes in certain characteristics of the image. Research shows that data hiding in the image will introduce the very rich random texture of high frequency component and also cause image noise features change. Thus, a new steganalysis blind detection method is proposed. Firstly, this method uses local linear transform to extract image texture features. Secondly, it extracts image noise features from three areas: wavelet analysis, image denoising and neighbor prediction. Thirdly, it calibrates all income characteristics to make them reflect the changes of embedded information better. Finally, it exploits support vector machine for feature classification to whether the image contains hidden information.Through testing the four typical steganographic methods: the LSB, Cox's SS, F5 and JPhide, the results show that this method can achieve blind detection analyses of hidden information effectively.