A Binary Classification Approach for Time Granular Traffic in SDWMN based IoT Networks

Wireless Mesh Networks (WMNs) constitute a major part of modern wireless Internet of Things (IoT) networks. Employing the SDN technology with WMN results in Software Defined WMN (SDWMN). The architectural independence of SDN allows to have heterogeneous and large number of wireless devices in SDWMN networks. This results in a mammoth amount of varying granular network traffic. To manage such massive and multi-granular traffic, various traffic engineering techniques are employed. Traffic Classification (TC) is one of such techniques which helps to classify this multi-granular traffic and draw the functional heuristics. TC proves to be beneficial in numerous application such as firewall building, intrusion detection, status report generation, etc. Time granularity based TC helps to achieve these objectives which result in improved network management. This paper performs the binary classification on time-granularity based IoT network traffic by employing different Machine Learning (ML) techniques. All of them achieve a reliable accuracy of more than 90%. Fine-KNN exhibits the best accuracy for most of the traffic classes with a rate of more than 98%. All classification accuracies are verified with cross-validation technique of 10-fold to prevent the chances of over-fitting. This verified accuracy assists to choose the suitable classifier in future network applications to assure the improved network performance.