Abstract: Model-based (MB) steganography for JPEG images can keep the histogram of DCT coefficients almost unchanged, therefore can resist the previous steganalytic methods. However, a loophole still exists. In MB steganography, a model derived from the original histogram is used so that the distribution of DCT coefficients in a smooth area becomes similar to that in a texture area. In view of this, we propose a steganalytic scheme against the MB steganography, which examines difference between DCT coefficient histogram in texture blocks and that of the whole image, and compare this difference with a threshold. Experimental results show that the proposed method can effectively detect the presence of secret data in an image embedded with the MB steganography.

Key words: JPEG image, modelbased steganography, steganalysis, information hiding