Secure bit-plane based steganography

Abstract

Steganography is the science of hiding information whose goal is to hide even the existence of secret messages by embedding the secret messages into an innocent-like cover. On the contrary, steganalysis is the science of detecting the existence of hidden messages in embedded image by steganography. Pixel-value differencing (PVD) steganography and bit-plane complexity segmentation (BPCS) steganography are reviewed. We analyze the weakness of these two steganography methods such as generating blocky effects and making noise in smooth areas of embedded image. Therefore, they are detectable with steganalysis methods. PVD steganography generates step effects or abnormal high fluctuations in PVD histogram. Recently, modified PVD (MPVD) steganography preserving the PVD histogram is recently presented. MPVD steganography seemed to be secure because it can avoid the weakness of PVD. In this thesis, we propose a novel steganalysis method for defeating MPVD steganography. We embed message one more time into the suspicious image to generate multiple features in image histogram and pixel value differencing histogram. These features are used to detect the images which are embedded by MPVD steganography. In our experiment, the proposed method is tested on more than 2,000 images. The result shows that our method can defeat the MPVD steganography method at high embedding rate. In addition, after studying about the weakness of PVD and BPCS steganography methods, we also propose a secure bit plane based steganography method on the spatial domain, which uses a robust measure to select noisy blocks for embedding messages. A matrix embedding technique is also applied to reduce the change of cover images. Given that the statistical property of cover images is well preserved in stego-images, the proposed method is undetectable by steganalysis that uses RS analysis or histogram-based analysis. The proposed method is compared with the PVD and BPCS steganography methods. Experiment...