Neural based steganography

Abstract

Steganography is the art of hiding messages inside a multimedia block. Data hiding can be done in different medias like text, images, and audio signals. It is very difficult for the hackers to find whether the media has any valid data or not. The goal of this research is to indirectly hide the data into a graphical image using a neural algorithm, as it adds additional complexity for the hackers. Our approach hides indirectly the secured binary bits along with some selected graphical image bits, based on the neural network algorithm, to get cipher bits. The generated cipher bits are then placed in least significant bit (LSB) position of the transmitted graphical image. In reverse process, this method regenerates the original data bits. The neural model used here is the multi-layer feed forward network. © Springer-Verlag Berlin Heidelberg 2004.