Image-based Steganalysis Technology Research

Steganography is a method of hiding data into other multimedia data.It was designed to provide a more secure data protection scheme than traditional encryption technology.Compared with traditional information encryption technology,steganography does not directly transmit secret data,but transfers data through the transmission of covers.So steganography can not only protect data,but also protect communication channel.In recent years,with the popularization of network and the development of computer technology,steganography has brought a certain security threat with its unique characteristics.It has been reported that the abuse use of steganography to transmit data,send Trojan horses,and steal information.Under this background,steganography,which is opposed to steganography,has rapidly become a research hotspot in the field of information security.In this paper,we both research active steganography and passive steganography,and the research results has been used in our DLP system.The Steganalysis model used in this paper is rich model which is used more in the present steganalysis due to its get better performance than the others.Faced with the security threats such as information leakage brought by steganography,and there are many drawbacks in the present steganalysis method,we proposed a method of active attack steganalysis method.Firstly we transform image into frequency domain through fast fourier transform,and filtering the high frequency data and low frequency data separately based on the influence on frequency domain caused by steganography.Then we filter the low frequency data with Wiener filter and the high frequency data with filter which are based on Markov model.Finally superimpose all the data in spatial domain.Experiment results show that this method can remove steganographic data efficiently and broke the covert communication which based on Steganography.Rich model is one of the best Steganalysis method ever,but it still has shortcomings due to the dimension of the steganalysis feature is too high,and it reduces the efficiency of steganalysis.We reduce the steganalysis feature dimension by specially designed filter kernel.Also we use more effictive ensemble fisher linear classifier.Experiment result shows that it improved the steganalysis ability and make the performance more stable.Based on our research,we designed and implemented a media wash system,which can be used for corporate outbound control and prevent data leakage.The system not only supports image file cleaning,but also supports the cleaning of embedded images in office documents.Also the system support file type checking and verification.