Steganalysis Of LSB Matching Based Average Gray Co-ocurrence Matrix Of The Sum Features

With the development of network technology digital information can be convenient, quick, speedy transmission to every corner of the world. However the security of data transmission becomes a problem solved. The steganograpy technology was applied to do terrorist sabotages, transfer illegal information, destroy the world peace, harm innocent people's life safety by terrorist organizations. The steganograpy technology like a double-edged sword, mixed blessing. Steganalysis technology have a great significance to safeguard the state security and social stability.Based on image secret technology according to different types of image, the corresponding secret technology is diversite, for JPEG compression image, BMP image as carriers are researched. This paper mainly according to BMP image as the research object, around the space steganograpic method LSB matching and do steganalysis research. This paper main work are as follows:(1) Introduce the modern information hidden theory background and application, he theory of information hiding produce makes information security communication possible.The paper in view of the additive noise model steganographic LSB matching algorithm safety analysis, the paper analyses some classical LSB matching steganalysis methods.The results show that in the embedded rate is near 100%, WAM steganalysis method have the best performance, detection accuracy can reach more than 95%, and the improvement of the HCF method and ALE steganalysis method's detection accuracy rate of 80% or more. When embedded rates is lower than 20%, WAM detection rates is 70% or so, the other two methods'detection rate is at about 60%. When the embedded is in a low rate, three methods of detection accuracy rate is low, can increase the probability of miscarriage.(2) In this paper a kind of average co-ocurrence matrix of the sum features steganalysis method. According to the image between adjacent pixels with strong relevance and combined with co-ocurrence matrix can reflect the pixels and its place of the relationship between characteristics, structure 8 d average co-ocurrence matrix of sum features. The experimental results show that the proposed method is better than classical method's detection accuracy. According to different embedded rate, compared with WAM steganalysis method can increased a lot, than improved HCF and ALE method can increased 15% or so. (3) Through the front experimental results, it can be found:when the stego image with the embedded rate near 100%, detection accuracy rate of some methods are near 95%, this can accurately detect the hidden algorithm. And when embedded rates is lower than 25%, the average detection is below 70%. This makes low rate of embedded steganalysis detection accuracy reduced. In order to solve the low rate of embedded detection with poor performance problem, by strengthening the different between the carrier image and stego image, with the aid of image restore carrier images obtained corresponding estimates that image, to be measured respectively from the image and estimate the image extracted 16 d average co-ocurrence matrix of the sum features, use the FISHER linear classifier classifies detection. Experimental results show that the embedded rate of 25%, detection rate is 94%; Even if the embedded rate was 10%, the detection rate of about 80% arrived, than did not use the way of dealing with image estimation increased by 10% or so.