Kernel Density Estimation On Correlated Naive Bayes Network Traffic Classification

Today Internet profoundly affects all aspects of people’s lives, the network has brought to the work and life more and more convenient, at the same time also caused many security and management problems. Network traffic classification is the foundation of network management, and the precondition of detecting or blocking malicious attacks and abnormal.With the rapid growth of network applications, the efficiency of traditional traffic classification was reduced greatly. By researching various types of network classification method, we found that most traffic flow classification methods do not consider the correlation among the flows at present. In this paper, flow correlation information is modeled by bag-of-flow (BoF), and a method of nonparametric kernel density estimation that applied to the Bayes classification is introduced. Also the correlated naive Bayes network traffic classification with kernel density estimation process is introduced and a software was coded for the method. The Bayes classification method is improved by performing the nonparametric kernel density estimation to various types of distribution functions. A real network traffic data is used in the experiment and other classifications are analyzed in this paper. The results of experiment achieve a better classification accuracy.