Detecting LSB Steganography Based on Noise Function

This paper presents ND (Noise functiondetection) steganalysis algorithm to detect LSB steganography. The purpose of constructing the noise function isto quantify the smoothness or "regularity" of the images.ND method is based on the property that LSB embeddingwill increase the noise functional value of the image, andthen the LSB embedding message ratio is estimated byconstructing the simple line equation with the statistics ofnoise function in image. Experimental results show thatthis algorithm is more accurate and has a lower missingrate and false alarm rate than the conventional RS. Compared with RS method and some other powerful steganalysis approaches presented recently, ND method directly usethe noise function to estimate the LSB embedding messageratio, neither fixed nor dynamic mask is needed. Thus,more running time is saved. The ND method is relativelyfaster, simpler and has good detection result.