Steganalysis of ±k steganography based on noncausal linear predictor

Abstract

The paper proposes a novel steganalytic technique for ±k steganography based on noncausal linear predictor using prediction coefficients obtained from the autocorrelation matrix for a block of pixels in the stego-image. The image is divided into equal-size blocks, autocorrelation matrix is found for the block, and the appropriate noncausal linear prediction coefficients is selected to predict all pixels in that block. A pixel is assumed to be embedded with message bit if the absolute difference between the original pixel value and predicted pixel value exceeds the pre-defined threshold. The effectiveness of the proposed technique is verified using different images.