Abstract:
Fog Computing is a decentralized technology that can execute and process data regionally and
can function on different systems, making it perfect for Internet of Things (IoT) applications.
According to a study by Gemalto in 2019, it has been found that about 52% of the companies
cannot even detect the IoT data breaches. Hence, The Network Intrusion Detection System
(NIDS) is an indispensable part of every fog and IoT security application and provides quality
of service. The number one safety venture is to form a technique for discovering intrusion
effectively and reduce the effect of it rapidly. However, because of the useful resource barriers
of fog and IoT devices, a light-weight IDS is fairly demanding. In this project, we propose a
NetFPGA based IoT Network Intrusion Detection system for fog computing architecture that
uses a two-layer model. Layer-1 model built on flow-level statistical features of the IoT
network, classifies the network flow on the type of application-layer protocol, while the layer2 model trained on flow-level features, detects intrusion in IoT networks. Our proposed
intrusion detection model, first categorizes the network flow as benign or malicious, then
classifies the category or subcategory of detected malicious activity. Our methodology inspects
packet headers to classify the network traffic in real-time on NetFPGA and uses flow-level
features extracted from the IoT-23 Dataset. The decision tree classifier yielded the highest
predictive results for both layer-1 and layer-2 i.e. an accuracy of 99.39% and 99.7%
respectively. This design has laid a solid foundation for the advent of an intrusion detection
system for the Internet of Things network, which will be of interest to academic and industrial
researchers.
