Research On Detecting Of Audio Steganography Techniques

Over the past twenty years, steganography has provided plenty of new approaches and challenges for information security development. Being the opposite and the critical technique to prevent it from being abused, steganalysis has been drawn tremendous attentions in an increasing demand. The technique of stego-detection has been seen as both the steganalysis basis and the preface of steganography extracting and attacking. Therefore the research on stego-detection is critical and significant to national security and social stabilization.In this thesis, based on the analysis of current audio steganography and steganalysis situation, both a universal and a special stego-detection algorithm are approached. Then a DCT differential statistical distribution of audio is studyed. The main contributions of the thesis are as follows.1. A study has made on an audio universal stego-detection algorithm, based on the analysis of main changes by audio steganography, a feature selection scheme of principal component analysis (PCA) is proposed. The feasibility of the PCA feature is monitored both in thesis and experiments. Then this feature method scheme is implemented by using support vector machine to classify, which proved to be more sensitive and accurate and also well performed in little imformation hiding.2. A study has made on a statistics stego-detection algorithm by audio LSB insertion. Based on the intrinsic behavior caused by audio LSB steganography, the change of audio adjacent statistical features is discussed. Then a feature selection scheme is approached with the neighboring feature. The feasibility is proved to be more adaptable, accurate and time expended.3. A research on DCT differential statistical distribution has developed. On the discussion of goodness-of-fit tests by x~2 methods, the experiment results have given an argument for inferring the DCT differential statistical distribution to have a Generalized Gaussian distribution. Then some forecasts in audio steganography and steganalysis application are forecasted.Finally, some further research directions in audio stego-detection techniques are prospected in this thesis.