Steganalysis of two least significant bits embedding based on least square method

The purpose of steganalysis is to detect the existence of the hidden messages. In this paper, a steganalysis algorithm against two least significant bits steganography is proposed in order to estimate the hidden message length in images. In the method, firstly we estimate an approximately original image through a local masked estimation function. Then we construct a weighted stego image by extending the definition which was proposed in LSB steganalysis to 2LSB. Finally, detection equation is formulated as a simple optimization problem between approximately original image and weighted stego image. Hidden message length can be easily calculated by solving the detection equation. Experimental results reveal that this algorithm can estimate the length of secret message with high accuracy.