Steganography And Steganalysis For Binary Images

With the acceleration of the global information digitization, numerous binary images are widely used in everyday life because of the convenient storage and effortless access. Therefore, hiding data in binary images for annotation, authentication, and copyright protection is in great demand. However, data hiding in binary images is a challenging task because of the lack of redundancy in the carrier. Arbitrarily flipping the pixel values can cause noticeable noise.In this dissertation, with a detailed survey of the currently available data hiding methods, we have carried out some exploration on data hiding in binary images. The main work is summarized as follows:1. Unlike the previous block-based methods for data hiding, we propose a novel method for hiding a relatively large amount of data directly in the boundary pixels of a binary image. Because all secret data are inserted on boundaries, good invisibility is assured. The central part of the scheme is the two sets of rules, one for data dependent embedding, and the other for data independent embedding. These rules are designed to ensure correct identification of the embedded information in data extraction. In other words, the eligibility status of the boundary pixels is not affected by data embedding therefore the original carrier image is not needed in extraction.2. We introduce a high capacity reversible data hiding method for binary images. This method is based on the correlation between the boundary pixels and its immediate neighbourhood. Each boundary pixel satisfying certain conditions is assigned a flag bit. It is shown that the sequence of all flag bits contains more zeros than ones therefore is losslessly compressible. The created space by the compression is used to carry secret message bits. The original binary image can be restored without error while the embedded information is extracted. Experimental results demonstrate validity and feasibility of the proposed method.3. A steganalytical scheme is proposed against block-based data hiding in gray images, in which the change in statistical characteristic caused by information hiding is used to detect the presence of secret message. Block-based data hiding procedure manipulates pixels of each block randomly and always introduces blockiness abnormality, and thus can be detected. This method is extended in the thesis to analyze other block-based data hiding schemes for binary images. Simulation results show that these methods are highly effective especially when the image size is not too small.