Research On DS Evidence Theory And Its Application Based On High Conflict Evidence Revision

With the network improving in recent years,network data has attracted more and more attention.All types of network data has been generated every day,and more and more problems have aroused in informational society.How to find anomalies in these huge data is an issue that we urgently need to solve.With network forensics technology came and rapidly developing.However,many problems were often encountered in the process of evidence fusion,information has the characteristics of mass quantity,rapid incensement and large quantity of emerging new words.In this paper,a BP neural network classifier model was proposed to solve the problem that a large amount of information.Network data collected was preprocessed to reduce the dimension of the fused data element.A D-S evidence fusion method based on evidence of high conflict correction was proposed and applied to the evidence fusion module of the forensic system.The work done in this paper can be summarized in the following three aspects:(1)Proposed an improved BP neural network classifier model.The algorithm mainly considers that the traditional BP neural network algorithm was easy to fall into the local optimum and had a bad influence on the classification result.By adjusting the dynamic factor learning method automatically,an improved BP neural network model is constructed,and applied it to the classification.Experimental results show that the algorithm can solve the problem of falling into a local optimum,it improved the efficiency of classification.(2)Proposed a fusion algorithm for correcting high-conflict evidence.This algorithm deeply analyzed the prone problems in the fusion process of D-S evidence theory.It combined the evidence's degree of trust and falsity to judge the high-conflict evidence,and corrected the high-conflict evidence to reduce the impact of high-conflict evidence on the fusion result.Through the experimented verification,this algorithm can solve the problem that the fusion result was contrary to the fact in the presence of high-conflict evidence very well,and improved the convergence speed.(3)Designed a network forensic evidence fusion module based on D-S evidence theory to preprocessed and analyzed network data flow.The BP neural network was combined with the D-S evidence theory and applied to the evidence fusion module of the network forensics system.In this system,BP neural network was used to classify the collected network data and generate meta evidence.Then the preprocessed evidence was fused.A reasonable consistency evidence was obtained and a forensic report was generated.The results confirmed that this algorithm could analyze the network data and determined the network behavior effectively.