Research Of Image Steganalysis Using Neighborhood Node Degree

With the rapid development of network communication technology, digitalmultimedia data become increasingly common. Digital media can be quickly andeasily transmitted through the communication network also led to the spread ofsteganography technology. Steganography is the art of covert communication byinvisibly embedding message into innocuous looking multimedia data, which can beused to transfer secret messages. Government departments, military agencies,financial institutions and other key sectors involved in the people’s livelihood takeadvantage of the data embedding technique to achieve secure transfer of confidentialdocuments. However it’s can be also utilized by the illegal organizations to transmitthe injurious information, which harms the social security and stabilization. Therefore,the detection of image hidden message, which is an opponent to steganography,possesses great significance.In this thesis, we concentrate on digital image, and the research includes detectionof least significant bit (LSB) matching steganography using neighborhood nodedegree and analysis the application of neighborhood node degree characteristics onspatial domain image steganalysis. The main contributions of this thesis are listed asfollows.(1) Spatial domain LSB matching can be modeled as adding independent additivenoise to the image. Natural images have a strong correlation between adjacent pixelsand it’s disturbed by LSB matching. Accordingly the effects of LSB matchingsteganography on neighborhood node degree are examined at first. Then usingsecondary steganography and two-dimensional histogram technique features areextracted from neighborhood node degree histogram. Support vector machine (SVM)classifiers are trained and tested with large cover and stego image databases.Experimental results testify that the proposed method possesses reliable detectionability and has superior results compared with other recently proposed algorithms.(2) Steganography disturbs the dependence between neighboring pixels, anddecreases the pixel’s neighborhood node degree. Based on neighborhood node degreecalibration algorithm, this thesis propose a new blind steganalysis against spatialdomain image. Image is calibrated using neighborhood node degree to generate acalibrated image at fisrt, then features are extracted from both test and calibratedimage on the dependence between neighboring pixels and image histogram. SVM isutilized to learn and discriminate the differences of features between original andstego images. Experimental results on large scale cover image and stego image testify that the proposed blind steganalysis method has a better performance.