Research Of Specific Testing Sample Steganalysis For JPEG Images

With the continuous development of information technology and communication technology,Internet communication has become an important way for people to trans-mit information.While enjoying the convenience brought by the Internet,how to com-municate safely and reliably has become a problem that needs to be solved in the field of network information security.The digital steganography technology embeds secret information in other digital carriers and transmits the carrier on the common channel to achieve the purpose of covert communication.But criminals also use the steganography to engage in illegal activities and endanger public safety.Steganalysis is a technology that aims at detecting embedded carriers generated by steganography.It is an important method to counter steganography and can provide security for network.There are many kinds of optional carriers for steganography.The steganography based on digital images is the most widely used and most advanced technology.The most common digital image carrier currently circulating on the Internet is the JPEG format,and most of the steganographic software uses the JPEG image as the carrier im-age.As the difficulty of applying JPEG steganography is reduced,in order to prevent lawbreakers from using steganography to engage in illegal activities and maintain net-work information security,how to effectively steganographically analyze JPEG images in existing network environments is a hot issue in the field of information security.In a real network environment,the source of digital images varies.The problem of"Cover Source Mismatch" is often encountered in steganalysis and the credibility of the steganalysis results is also unknown,which has brought a lot of difficulties to the conversion of the steganalysis to the real scene.This thesis presents a steganalysis framework that can be effectively used in big data real scenarios.The framework is based on existing steganalysis features and is suit-able for steganalysis using feature trained classifiers.Because the information hiding technology based on JPEG images is more advanced and JPEG images are widely used in the network.This thesis takes JPEG image steganalysis as an example to conduct an in-depth study of "Specific Testing Sample Steganalysis(STSS)".The main research contents and innovations of this thesis are as follows:1.Study the Influence of Steganographic Operations on Steganalysis Features The digital image steganography has made a slight modification to the image itself.If such a modification can be successfully recognized by the classifier,the correspond-ing steganalysis feature of the image must also have been changed.In order to determine the changing law of the above phenomenon,this thesis studies the influence of stegano-graphic operations on the image in the feature level.By observing the steganalysis of various angles,a series of patterns of image feature changes caused by steganography were found,the causes of CSM problems were found out,and a mitigation plan was proposed.2.Propose Method of Specific Testing Sample SteganalysisFor the problem that the steganalysis method is difficult to maintain high precision in real scenes,this thesis has found an important relationship between steganalysis and image features under the environment of big data training resources.A steganalysis method for selecting a specific training set for a testing sample is proposed-"Specific Testing Sample Steganalysis".This method defines the similarity between images,cal-culates the similarity measure between the test sample and the big data training set data,uses the similarity data as the training set of the test image,effectively solves the CSM problem and improves the accuracy of steganographic analysis.The classical JPEG steganography algorithm nsF5 and the mainstream JPEG steganalysis features,such as CC-CHEN,CC-PEV,C.F*,DCTR and GFR,are used as an example to conduct the experiment.The results show that the accuracy of this method is higher than that of existing methods.3.Propose Calculation Method of STSS ConfidenceFor the existing steganalysis methods,the credibility of the judgment result is un-known.This thesis proposes a steganalysis confidence calculation method based on STSS,and designs a confidence calculation neural network.The network integrates test samples,training data,classifiers,embedded rates,and other information.The confi-dence network can obtain confidence corresponding to error rate of steganographic anal-ysis results.The classic JPEG steganography algorithms nsF5 and steganalysis method CC-CHEN are used to conduct experiments.The results show that the confidence gen-erated by the CCNN can be used as reference information for analysts.