Skip to Main Content
UPnP Forum specifies a framework for quality of service (QoS) in its UPnP QoS architecture. The framework enables automatic configuration of priority based QoS to network devices according to local policies. The downside of the framework is that it relies on a control point, which acts as an initiator for UPnP QoS operations. In practice, this means that each application should explicitly support UPnP QoS to be able to use the services provided by the framework. In this paper, we propose a supplemental component for the architecture which enables the use of UPnP QoS without any support from the applications using the network. The network traffic is listened by a passive component which uses statistical methods for classifying the network traffic flows and acts as a control point for the UPnP QoS framework. Our results show that even very simple behavior-based methods can be used for traffic classification with a fairly good precision. In our experiment, the automatic UPnP QoS control point was able to perform the classification with an overall accuracy of 92% based only on the packet sizes of the four first packet of each flow. We assume that the accuracy can be further improved using additional mechanisms after the initial fast classification.