Broadband network traffic characterization and classification using a multivariate statistical method

Network traffic behavior is constantly changing due to issues as high service demand on a given service, network attacks, emergence of new services, among others. Although network traffic characterization and classification is a well-known task, it must mainly be effective in real-time anomalous situations in order to help to keep the network with good performance. Classical approaches such as the use of artificial intelligence mechanisms have been modified in order to attempt requirements. However, in general, such adaptations are slow, need of many resources and a constant participation of the network administrator. This work presents an investigation of a methodology for characterize and classify patterns into broadband network traffic. This methodology is based on flow clustering analysis, a multivariate statistical method employed for discovering associations and structures in collected data. The clustering analysis enables the extraction of patterns of data flows.