NSVM: A New SVM Algorithm Based on Traffic Flow Metric

Abstract

Support Vector Machines (SVM) is one of the most popular Machine Learning (ML) tools that can be applied to the problem of traffic classification in IP networks. For SVMs, there are still some open questions that need to be solved before they can be generally applied to traffic classifiers. Having being designed essentially as techniques for binary classification, their generalization to multi-class problems is still under research. Furthermore, their performance can be highly impacted by the correct optimization of their working parameters. In this paper we propose an approach to traffic classification based on SVM. We apply it to solving multi-class problems with SVMs to the task of statistical traffic classification, and describe a simple optimization algorithm that allows the classifier to perform correctly with as little training as a few hundred samples. The accuracy of the proposed classifier is evaluated over two sets of traffic traces, coming from different topological points in the Internet. Important parameter about SVM is deeply analyzed. The impact of packet sampling on traffic classification is studied. Experiment results show our proposed NSVM method can effective improve the classification results and reduce training set sizes. We can apply the method into network management with sampling technology.