JPEG Steganography with Dual Histogram Compensation using Theoretical Estimation Techniques

Steganography is the art of secret communication between two parties that not only conceals the contents of a message, but also its existence. Steganalysis attempts to detect the existence of embedded data in a steganographically altered cover file. Many algorithms have been proposed, but so far each has had some weakness that has allowed its effects to be detected, usually through first or second order statistical analysis of the image. Our algorithm, named J4, is a new steganographic technique that improves on our previous system, J3. J4 uses dual histogram restoration along with matrix encoding to minimize detection by first as well as second order steganalysis, whereas J3 only restored the global histogram. J4 embeds data efficiently by theoretical estimation of the optimal value of matrix encoding and payload for each coefficient pair, which is also done by the receiver, so that the stopping points do not need to be included as they were in J3. Existing blind steganalysis tests performed on J4 show that it has a detection rate of 50-60% at 0.05-0.1 bit per non-zero coefficient (bpnz), as compared to 80-90% for F5, OutGuess, Steghide, MB1, MB2 and PQ. Its performance is comparable to the best known algorithm, nsF5, slightly outperforming it at embedding rates of 0.1 bpnz and higher. Since 50% detection rate is classified as random guess using an SVM classifier, J4 is an ideal candidate for a steganographic algorithm.