A differential evolution based algorithm for breaking the visual steganalytic system

Image steganography is the process of sending messages secretly by hiding the message in image content. Steganalytic techniques are used to detect whether an image contains a hidden message by analyzing various image features between stego-images (the images containing hidden messages) and cover-images (the images containing no hidden messages). In the past, genetic algorithm (GA) was applied to design a robust steganographic system that breaks the steganalytic systems. However, GA consumes too much time to converge to the optimal solution. In this paper, we use a different evolutionary approach, named differential evolution (DE), to increase the performance of the steganographic system. The key element that DE is distinguished from other population based approaches is differential mutation, which aims to find the global optimum of a multidimensional, multimodal function. Experimental results show that the application of the DE based steganography not only improves the peak signal to noise ratio (PSNR) of the stego-image, but also promotes the normalized correlation (NC) of the extracted secret message for the same number of iterations. It is observed that the percentage increase in PSNR values ranges from 5% to 13% and that of NC values ranges from 0.8% to 3%.