One of the biggest challenges in universal steganalysis is the identification of reliable features that can be used to detect stego images. In this paper, we present a steganalysis method using features calculated from a measure that is invariant for cover images and is altered for stego images. We derive this measure, which is the ratio of any two Fourier coefficients of the distribution of the DCT coefficients, by modeling the distribution of the DCT coefficients as a Laplacian. We evaluate our steganalysis detector against three different pixel-domain steganography techniques.
展开▼
