The Research On Reversible Data Embedding Techniques

With the digitalization of data and communications via network, secret protection of communications over the Internet is becoming more and more crucial. Several different approaches encrypt secret messages by transforming them into an unrecognizable form. However, meaningless encrypted messages can attract the attention of attackers during transmission. To remedy this, information hiding methods embed secret data into cover objects such as written texts, digital images, and videos in order to make confidential data imperceptible.This dissertation includes three kinds of information hiding techniques: reversible information hiding schemes in VQ-compressed domain, secret sharing schemes for digital image,information hiding schemes in text and optimizing least-significant-bit substitution method.As to the reversible information hiding schemes in VQ-compressed domain, we propose three new methods to improve the hiding capacity and the compression rate. The first one is a reversible information hiding scheme for VQ-compressed images using prediction errors. This scheme exploits characteristics of VQ indices and the prediction errors between them to embed the secret data and to guarantee that the original indices can be recovered to reconstruct the VQ compressed image. We also propose a data embedding scheme with reversibility for VQ compressed images based on index lookup technique. The third scheme aims to minimize the size expansion of the embedded result. To achieve this goal, the index appearance frequency histogram was analyzed before the embedding process started. We compare our methods with Yang and Lin's method, Wang and Lu's method, Chang et al.'s method proposed in 2009 and Chang et al.'s method proposed in 2007. The experimental results show that the first proposed method can achieve the best bit rate 0.5bpp while comparing with other schemes. Our second proposed scheme yields greater hiding capacity about 48000 bits when applied in the smooth images sized 512×512. The third method achieves good hiding capacity while maintains lower compression rate.There are two new secret image sharing schemes proposed in this manuscript. The first one is a verifiable secret sharing scheme for binary images. This scheme uses a watermark image to verify the reconstructed secret image so that a defined participant does not need to execute shadow verification during both the shares reconstruction phase and the revealing phase. The second proposed scheme derives the secret shadows and generates the meaningful shadow images by adopting the sudoku. The experiments show that the shadows can be successfully camouflaged in the host image with satisfactory quality. The distortion of the embedded host pixels is limited within range [0,3]. Moreover, the proposed scheme provides a large capacity for embedded secret data.In addition, we propose two new text information hiding schemes and one new optimizing least-significant-bit.subtimization method in this dissertation. The reversible information hiding scheme using left-right and up-down Chinese character representation improves Sun et al.'s L-R scheme that makes the embedding and extracting procedures for Chinese characters very simple and render advance transformation of the text file into an image file unnecessary. On the other hand, the emoticon-based text steganography in chat is proposed to embed secret information into emotional icons (also called emoticons for short) in chat rooms over the Internet. By using this technique, the receiver can extract the secret while the cover text does not have any distortion. The optimizing least-significant-bit subtimization method adopts the cat swarm optimization (CSO) strategy to obtain the optimal or near optimal solution of the stego-image quality problem. The experimental results show that this scheme can obtain a better solution with less computation time.