Abstract: Currently, image steganalysis has evolved into a significant subject of interest in the field of information security. In accordance with the region stationary characteristic of the image source, a new steganalytic method that can exploit the image region randomness features is proposed based on analyzing the changes of the regional randomness caused by the least significant bit (LSB) matching steganography. First of all, a given image is separated into a great many of block regions, and the regional LSB sequences are extracted from the pixel sequences of each block region which are obtained by the Hilbert scanning. Furthermore, the parameter which is derived from a LSB sequence with the bit XOR operation, is defined as the region randomness measurement index. Finally, the histogram of the index is calculated and analyzed; three sorts of histogram features including the information entropy, the special histogram values, and the moments are extracted; and the Fisher linear discriminator is applied for classifying the cover images and stego images. Extensive experimental results on three grayscale image databases with different embedding rates show that the proposed approach has good performance in detecting LSB matching steganographic scheme, and it obviously outperforms the previous typical steganalytic algorithms.