Research On Network Traffic Classification Based On Bayes Theory

As the network technology develops fastly, especially the rapid development of P2P applications, which greatly abundances the content of the internet. Most of their services use port-hopping and protocol- encrypted and other related technologies, the traditional classification efficiency based on port and payload greatly reduce. How to classify network traffic effectively and providing the vast number of internet users with a safe, reliable and efficient use of the environment, which is an important problem for a lot of scholars and network management. Taking use of machine learning method for automatic classification of network traffic is an effective way. Topics research is concern of Network traffic classification based on Bayesian threory.First of all, topics introduces the collection method of network traffic data, it takes use of Wincap software to capture network packet, then the packets collected are classified to the traffic by 5 tuple, results in 34 candidate characteristics of network traffic and forms a network traffic feature vector. On the tagging of the network traffic applications, it combines the port-based and machine IP approach to classify the network traffic, and then forms traffic of samples.In network traffic feature selection, topics introduces that it makes use of correlation and fast filters selection method to select features, find the characteristics of a subset which is propitious to classification of the newwork traffic, the results of experiments shows that the raised method can not only drop the characteristic dimension in order to reduce the time of study and classification, but also remove irrelevant or redundant features, increase the accuracy of classification.In aspect of network traffic classification,we analysis and do some researches on the Naive Bayes, Bayesian network, C4.5 decision tree classifier methods, which led to the construction of three networkat traffic classifier, it uses 8 kinds of mixed common application types of network traffic data for the classification study and give the experimental results, which show that the research based on Bayesian classification methods of network traffic classification accuracy is high,as well as the classification time is few , it has some application value With the above research.