Software Defined Machine Learning Based Anomaly Detection In Fog Based Iot Network

Adaptation of intelligent devices with smart connectivity has tremendously increased the Internet of Things (IoT) traffic. For the security of IoTs massive applications, anomaly detection in these large number of devices is resource intensive task. This anomaly detection neither can be done in a cloud where analytical applications run nor in IoT device due to its limited computation capability. The Software Defined Networking (SDN) promises better management of network due to centralized control. In this paper, we proposed a software-defined machine learning (ML) based anomaly detection framework that provides network control in two scenarios, first is cloud infrastructure, and other is collocated fog(network edge) infrastructure. We compare our work in terms of delay in cloud against a collocated fog (processing) node; furthermore, we have also evaluated both scenarios with respect to packet error rate and throughput. Moreover, we discuss that these factors are critical in attack detection and mitigation. In the end, we conclude that in machine learning based anomaly detection, fog nodes provide better computational (attack detection) results in comparison with a cloud infrastructure.