A new image steganography technique based on similarity in secret message

In this paper, we propose a new image steganography scheme for colored images based on the cluster analysis. In this scheme, we analyze the secret data in order to make its clusters. The secret data can be textual, image/video or audio/speech. A cluster contains ASCII values of characters if the secret data is text, sample values for audio/speech and pixel values in case it is image/video. We then calculate the difference value between the secret data and the minimum value contained in the cluster. We do not hide actual secret data; the difference value is embedded equally into two channels of the image. The experimental results show that our proposed method has enhanced security as compared to the modified Kekre algorithm [7] and pixel intensity based high capacity data embedding method [9]. Furthermore, our scheme has good hiding capacity, high PSNR value, and very low MSE value.