Blind decoding of image steganography using entropy model

Abstract

Image steganography hides secret information in the cover images so naturally that the existence of hidden data in the stego-image is not recognisable. This Letter proposes a new approach to blind decoding of image steganography using the local entropy distributions of decoded images. The local entropy distributions of incorrectly decoded images are different from those of normal ones because of the abnormal image structures in the erroneously decoded images. This blind decoding in the image steganography is very useful to extract hidden image information because there are enormous least significant bit (LSB)-based steganography methods, and it is very hard to find the methods by observing manipulated LSBs.