Classification of LSB and MLSB stego images via shift and pixel pairs irrelevance

For distinguish the LSB (Least significant bit) replacement stego image from MLSB (Multiple least significant bits) stego image, which are two typical kinds of steganographical methods of image spatial domain and have been applied widely, a classification algorithm based on the shift of pixel value and irrelevance of pixel pairs is proposed. In this algorithm, a shift operator is adopted for each pixel value of test image, and then the irrelevance of a kind of special adjoint pixels is extracted as the feature, at last the BP (Back-Propagation) neural network is designed to classify LSB replacement and MLSB replacement images. Results of experiments show that the proposed method can distinguish the MLSB replacement stego images from LSB replacement stego images reliably.