Abstract: The research on steganalysis has mainly focused on hidden information detection, and there are few methods about active steganalysis. From the view of hidden information detection, the pixels are classified into different kinds. Based on the analysis of the effects on the frequencies of different kinds of pixels by message embedding and Least Significant Bit (LSB) plane flipping, an active steganalysis approach is proposed to recover the stego key of LSB replacement steganography in spatial domain of images. This method has remarkable physical significance and can be implemented conveniently. Experimental results show that this method can recover the stego key successfully in certain range of embedding ratio.