Study On Image Steganalysis And Its Evaluation Method

As the fast development of network and communication technique, steganography as an important technology in information security has attracted high attention of many scholars and related departments.Recent years, according to steganography which is covert communication means,reverse steganography-steganalysis has became hot issues in information security field.However because the massive proposed steganalysis algorithms are so different that it is difficult to choose a proper one for practical application.So how to effectively evaluate various steganalysis algorithms'performance has became a problem which is urgent to be solved.At present the most existing steganalysis have the disadvantages of low detection rate,by deep study the basic theory and related technology of steganalysis,a new blind detection method of JPEG image is proposed, in which DCT coefficients of JPEG images are separated into wavelet domain and then a series of higher-order statistics features of the sub-band wavelet coefficients are calculated,the feature sealing was introduced and it makes the improved method more sensitive to tiny feature difference, then the classifier based on Support Vector Machine is applied to finish the reliable detection of image.The results show that the proposed method improved the correct detection rate and verifies it is an effective steganalysis.Besides, at present the most existing steganalysis evaluation method have the disadvantages of calculation complexity and irrational weight distribution in the evaluation process.In consideration of these characteristics the paper combines grey correlation degree theory with entropy weighted method and uses weighted grey correlation analysis to calculate the weighted grey correlation degree between the target index vector and optimal index vector, the problem of steganalysis'reliability evaluation is resolved with this kind of method.The instance simulation proves that the proposed method can reasonably evaluate the steganalysis algorithm and verifies it is a feasible method.The improved steganalysis of higher-order statistics in wavelet domain is evaluated using weighted grey correlation analysis methord, the evaluation results prove the effectiveness of the evaluation method once again.