Digital Image Watermarking And Steganalysis

Digital watermarking has become a vibrant research area of information hiding over the last several years. It can embed secret information into redundant bits of digital media contents to hide its existence. Different from encryption, information hiding or steganography is to hide the very presence of communication and therefore it is more covert and secure than encryption. With the development of Internet and multimedia technology, it is possible to communicate secretly via Internet using information hiding, which might compromise the information security of our country.Based on statistical properties, steganalysis is the art of discovering hidden data in cover objects, which is the key step of revealing secret communication.In this paper, we first study some basic concepts of digital watermarking and its application on information hiding, and further investigate the detection of steganography, including the feasibility of steganalysis and the method of developing steganalytic algorithms. After that, some practical steganalytic algorithms are proposed and experimental results show that these algorithms are capable of discovering the existence of hidden data in digital images.On the basis of RS algorithm, Quadratic Regular-Singular Steganalysis (QRS) is proposed by improving the fitting model. Simulation results show that the improved method is more accurate and less time-consuming than other existing algorithms. And therefore we conclude that QRS is an effective method for detecting LSB embedding in digital images.JPEG format is the most popular compression format for delivering and storing digital images due to its efficiency in data storage, quality of compressed image, and its flexibility. It is inevitable to use images that were originally stored in the JPEG format as cover images in steganography. As JPEG compression is a kind of lossy compression, it leaves unique fingerprints and serves as a fragile watermark, which enables detecting small changes of the cover image. Steganalysis based on JPEG compatibility is capable of detecting secret messages embedded in images that were originally stored in the JPEG format.In the last part of the paper, we study the concept of universal steganalysis. Universal steganalysis can discover secret messages embedded by any steganographic tools as long as it is trained properly to learn the differences between cover and stego image features. Because of the potential of discovering unknown steganography, universal steganalysis is of great value in practical application. After we carefully studied all the feature vectors used to train a support vector machine (SVM), a method of calculating feature vectors from different parts of the image separately according to their texture is proposed and the extracted feature vectors are also improved to increase the efficiency. Our future efforts will focus on improving the performance of universal steganalysis using these methods.