Performance improvement of LSB-based steganalysis using bit-plane decomposition of images

Abstract

In this paper, we present an improved least significant bit (LSB)-based steganalysis scheme using the bit-plane decomposition of images. We derive a mathematical condition that can enhance the detection rate for hidden messages based on the correlation coefficient between two parts of a decomposed image. Based on this condition, images are classified and segregated into two groups: the full image including all of the bit-planes and a sub-image containing only the lower bit-planes. The feature vectors for steganalysis are extracted independently form each group. Three types of conventional feature vectors were extracted to verify our proposed method and experiments demonstrated that conventional steganalysis schemes exhibited improved performance using our proposed method. In conclusion, our scheme can be used as a general steganalyzer regardless of the specific steganalysis methods employed for LSB-based steganalysis.