A Survey on Feature Selection Techniques for Internet Traffic Classification
In recent years, a technique based on machinelearning for Internet traffic classification has attracted more and more attentions. It not only overcomes some shortcomingsof traditional classification technique based on port number,but also does not inspect the packet payload, which involves the security and privacy. In this paper, we apply an unsupervised machine learning approach based on DBSCANalgorithm. DBSCAN algorithm has three merits: (1) minimalrequirements of domain knowledge to determine the inputparameters; (2) discovery of clusters with arbitrary shapes; (3)good efficiency on large data set. Experiment results show that DBSCAN has better effectiveness and efficiency.