Traffic classification using probabilistic neural networks

Traffic classification, a branch of passive network measurement, becomes more and more important for network management. As traditional traffic classification techniques like port-based and payload-based techniques become ineffective for complicated internet applications which use dynamic port number and encryption techniques to avoid detection, machine learning based techniques gained more and more attentions in the past few years. But there are few studies that focus on applying neural computation techniques for traffic classification. In this paper, we use a distributed host based traffic collection platform (DHTCP) to gather traffic samples with accurate application information on user hosts. Then probabilistic neural network was used to traffic classification. Web and P2P traffics were studied since they are the most predominant internet traffic types. experimental results show that probabilistic neural network is an effective machine learning technique for traffic identification.