Steganalysis of LSB-Matching steganography by removing most significant bit planes

This paper proposed a new steganalysis scheme of LSB-matching steganography based on statistical moments of the DFT of histogram of multi-level wavelet subbands. Before deriving these wavelet subbands a pre-processing apply to images under the test. The pre-processing contains removing some most significant bit planes. Then we decompose the image using three-level Haar discrete wavelet transform (DWT) into 13 subbands (here the image itself is considered as the LL0 subband).The Fourier transform of each subband histogram, is calculated. Then it is divided into low and high frequency bands. The first three statistical moments of each band are selected to form a 78-dimensional feature vector for Steganalysis. Support vector machines (SVM) classifier is then used to discriminate between stego-images and innocent images.