An Online Classification Approach Based on Universal Inter-arrival Time

In this paper we present an approach for online traffic classification based on the statistical analysis of protocol behaviour at IP level. Then we use the statistics of protocol attribute and Bayesian network method to build a classifier,which can classify unknown flows dynamically as packets pass through the classifier, deciding if a flow belongs to a given application. Distinct from other methods, we use the ldquouniverse inter-arrival timerdquo to overcome the influence of RTT variance so that the statistic of universe inter-arrival time is site-independent and time-independent. At last, the experimental results show that our approach performs better than other methods using tradition inter-arrival time.