Steganalsis Of Spread Spectrum And QIM Information Hiding In Image

The rapid development of multimedia and internet technology and recognition of individual privacy and digital copyright protection promote a rapid development in information hiding. Information hiding as a new information secure technology not only provides a solution for digital copyright protection but also can combine with cryptography to transit information secretly. But the bonus of indefectibility by using information hiding to communicate could be applied by terrorists to transfer illegal information, and the secure of the country and society may be jeoparded. Therefore, to detect the hiding information in multimedia which is steganalysis, gained much more attention on the governments, intelligence agencies and experts and scholars in the research area. Steganalysis becomes a hotspot nowadays. The steganalysis of information hiding algorithms based on spatial domain which are usually proposed many years ago, has enjoyed a good development. But the steganalysis of information hiding algorithms based on frequency domain which gain much more appliance and possess much more robustness and statistical imperceptibility, has not developed adequately so far. Aimed at the two popular frequency domain information hiding algorithms which are based on spread spectrum and QIM, the thesis proposes steganalysis algorithms with certain capability of generalization. To the spread spectrum based information hiding, the thesis applies Markov-chain to capture the relation between the nearby pixels, gains feature vectors by PMF matrix and employs two methods to classify. And to the QIM data hiding the thesis chooses to analyses the histogram of the DCT domain of the images. The main research work and contribution of this thesis are as follows:1) At first, the thesis expatiates on the origin and development status of information hiding and compares steganography with cryptography to show its excellence and specialty. And then this thesis introduces the spatial and frequency domain information hiding algorithms. At last systematically summarizes research status of special steganalysis of spatial and frequency domains as well as general steganalysis.2) The specialties and hiding process of spread spectrum and quantization index modulation information hiding are introduced particularly. The thesis systematically analyses several features that could be used by steganalysis, for example the changes of energy, relativity of the nearby pixels and information entropy after hiding data, and those provide certain theory ways for steganalysis.3) The thesis presents an algorithm for steganalysis of images based on spread spectrum data hiding. Employing the Markov chains, whether an image contains hiding data or not can be discriminated with combination of the threshold method and the improved pattern recognition method, through detecting the relativity of correlated pixels of the image. Experimental results based on Corel's image library show that this algorithm, by reducing the false alarm rate, improves the precision rate than before.4) To the steganalysis of QIM based data hiding, through analyzing the probability distributing stat. combined with the additive noise model, the thesis separates the DCT domain histogram as feature vectors. Through analyzing the characteristic of images with low QIM stego rate and the experiment data, add the derivation of the histogram as another feature vectors. The experimental results shows that adding the derivation of the histogram as feature vectors promotes the precision rate with low stego rate, and when the QIM step is determined the result is quite good.