LSB data hiding detection based on gray level co-occurrence matrix (GLCM)

In this paper we present a novel steganalysis method with feature vectors derived from gray level co-occurrence matrix (GLCM) in spatial domain, which is sensitive to data embedding process. This GLCM matrix is derived from an image. We consider several combinations of diagonal elements of GLCM as features and use SVM for classification. The experimental results have demonstrated that the proposed scheme outperforms the existing steganalysis techniques in attacking the LSB steganographic schemes applied to spatial domain. Our results also show that there are different between these features for stego and non-stego images and these features are convenient for steganalysis. With randomly selected 900 images for training and the remaining 900 images for testing, the proposed steganalysis system can achieve a correct classification rate of 98.1% for LSB (0.1 bpp) and 81.1 % for LSB matching algorithm. For combination of algorithms we reach to 95.6% correct detection rate.