Image Steganography for Pixel Prediction using K-nearest Neighbor

Abstract

: Nowadays to secure the privacy of the patient has increased more research interest during the Image steganography process. Least Significant Bit (LSB) substitute approach was widely exploited to hide the sensitive information in the conventional works. Here, each pixel was reinstated to achieve advanced privacy, other than it increased the complexity. This paper develops a new pixel prediction model-based image steganography to surmount the complication problems widespread in the conventional works. In the proposed pixel prediction model, the K-Nearest Neighbour (KNN) classifier is used to construct the prediction map that recognizes the appropriate pixels for the embedding process. Subsequently, from the medical image to extract the wavelet coefficients based on the Discrete Wavelet Transform (DWT) and embedding power and the undisclosed message is embedded into the HL wavelet band in the embedding phase. At last, from the medical image, the concealed message is extracted by using the DWT. The simulation of the proposed pixel prediction model is carried out by exploiting medical images from the BRATS database. The proposed pixel prediction model has attained maximum performance for the Peak Signal to Noise Ratio (PSNR), Structural Similarity Index (SSIM), and correlation factor, correspondingly.