Collaborative Flow-Identification Mechanism for Software-Defined Internet of Things

Due to the lack of unified standards in the Internet of Things (IoT), heterogeneity in terms of protocol and packet types exist. In such a case, accurate traffic identification using port-based and payload-based solutions are not suitable for flow processing in Software-Defined IoT (SD-IoT). In this article, we propose iAcceSD, an intelligent Access Node (SD-Access) and SDN controller (SD-Controller) collaborated traffic identification mechanism for SD-IoT. Concerning the heterogeneous and unknown traffic flows in IoT, iAcceSD uses machine learning (ML)-based traffic identification mechanisms at the access node. The SD-Controller trains a lightweight ML module for a specific SD-Access node dealing with a similar set of traffic flows. By processing flows at the edge, network latency is reduced in the proposed scheme. An optimization model is developed to program the SD-Access nodes considering the heterogeneous and unknown traffic. Thorough performance analysis of iAcceSD shows improvement in latency by 34% and controller overheads by 23.4%, compared to the existing state of the art, with a simultaneous improvement in energy consumption and packet delivery ratio.