Augmenting DiffServ operations with dynamically learned classes of services

In this work, we provide a Machine Learning framework for augmenting the Differentiated Services (DiffServ) protocol with fine-grained dynamic traffic classification. The is called L-DiffServ. It composed of two classification algorithms able to detect QoS classes incoming packets only looking at three packet header fields; first algorithm, referred as Inter-L-DiffServ, semi-supervised procedure replicate DiffServ classification; second one, Intra-L-DiffServ, an unsupervised algorithm intra-class classification, useful taking large portions overall traffic. We apply latter low priority best-effort class. performance evaluation shows that our solution dynamically classify and new sub-classes hence adapting aggregate characteristics. also show network resource management can be improved exploiting generated sub-classes: active queue based on WRED CHOKe reduction number sessions affected by losses up 40% respect legacy procedure.