A Lightweight Anomaly Detection Model using SVM for WSNs in IoT through a Hybrid Feature Selection Algorithm based on GA and GWO

As a result of an incredibly fast growth of the number and diversity of smart devices connectable to the internet, commonly through open wireless sensor networks (WSNs) in internet of things (IoT), the access of attackers to the network traffic in the form of intercepting, eavesdropping and rebroadcasting has become much easier. Anomaly or intrusion detection system (IDS) is an efficient security mechanism, however despite the maturity of anomaly detection technologies for wired networks, current technologies with high computational complexity are improper for resource-limited WSNs in IoT and they also fail to detect new WSN attacks. Furthermore, dealing with the huge amount of intrusion wireless traffic collected by sensors, causing slow detecting process, higher resource usage and inaccurate detection. Hence, considering WSN limitations for developing an IDS in IoT, establishes a significant challenge for security researchers. This paper proposes a new model to develop a support vector machine (SVM)-based lightweight IDS (LIDS) using combination concepts of genetic algorithm (GA) and mathematical equations of grey wolf optimizer (GWO) which is called GABGWO. The GABGWO through applying two new crossover and mutation operators tries to find the most relevant traffic features and eliminate worthless ones, in order to increase the performance of the LIDS. The performance of LIDS is evaluated using AWID real-world wireless dataset under two scenarios with and without using GABGWO. The results showed a promising behavior of the proposed GABGWO algorithm in choosing optimal traffics, decreasing the computational costs and providing high accuracies for LIDS. The hybrid algorithm is also compared to pure GA and GWO and other recent methods and it is found that its performance is better than them.