Effectiveness of CF Moments and Prediction Error Algorithm for Image Steganalysis

Covert Communication using digital images is rapidly gaining popularity. It alters some of the image properties that may introduce few degradation or unusual characteristics. These characteristics may act as signatures that broadcast the existence of the embedded message and thus defeating the purpose of steganography. With large number of techniques being developed in image steganography, universal Steganalysis has become essential. In this paper, effectiveness of statistical moments of wavelet characteristic function and prediction error image is discussed. Stego image has irregular statistical characteristics as compared to cover image. Bayes classifier is used to discriminate the cover and stego image. Database of 67 natural, food and animals´ images has been used for experimental work. This method has been tested against various popular steganography methods like Cox et al. spread spectrum technique, LSB generic, S-tool and F5 steganography algorithm.