Output details

Classification of Traffic Flows into QoS

Classes by Unsupervised Learning and KNN Clustering

This paper resulted from a U.S. National Science Foundation project on automatic classification of network traffic for adaptive resource management. The study showed that different traffic classes could be identified accurately just from their statistical behaviour without looking into packet contents. It also identified the most significant statistical features in traffic data. Unlike previous studies, this paper used self-organizing maps in unsupervised learning to identify inherent clusters in the data. This work has implications for law enforcement or intelligence faced with encrypted traffic where packet contents are concealed but statistical features can reveal the nature of the communications.