On The Improvement of the Tri-Way Pixel Value Differencing Steganography Algorithm.

Steganography consists of hiding information into digital files so that they cannot be noticeable to human vision. These files include texts, images, audios, videos and protocols. In this work we consider only gray-scale images as they are the most used digital covers in steganography. The most challenging problem consists of finding the right pixels of the image to embed or hide the maximum amount of information without deteriorating its quality. Research in this area is still at the premature phase even though several steganography image-based algorithms have been proposed recently. In this paper, we study the well-known tri-way pixel value differencing algorithm (TPVD) aiming at improving its performance. We select randomly the starting pixel of the pixel-pair combinations to hide information rather than starting always with the first pixel as done in the TPVD regular behavior algorithm. Our first experiment shows that a slight improvement can be obtained with these random selections in preserving the quality of the stegoimage (i.e., the image holding the information) and maximizing the amount of hidden information. We study also the encryption of the sensible information that should be embedded into the cover image using the AES algorithm. These randomness behavior along with the encryption technique render the retrieval of the hidden information very hard once the image is spotted as suspicious to be a stegoimage and the hidden information are attempted to be extracted. Our second experiment shows that the encryption of the same amount of information may deteriorate the quality of the stego-image and makes it somehow perceptible for human vision and also vulnerable to stego-analysis techniques. [ABSTRACT FROM AUTHOR]