Abstract:

A new steganalysis algorithm on model-based (MB) steganography is proposed after researching the high order statistics of images. The difference, imported by MB steganography is described by the co-occurrence discrete cosine transform(DCT) matrix, the boundary gradient energy in airspace and the imaginary eigenvalue decomposition. The MB steganography is detected using all of the features and support vector machine. Experiments show that this algorithm has a high accuracy rate of detection with a low false detection rate.

Key words: model-base steganography, imaginary eigenvalue decomposition, support vector machine