An Anomaly-Misuse Hybrid System for Efficient Intrusion Detection in Clustered Wireless Sensor Network Using Neural Network

Wireless Sensor Networks (WSN) refer to a group of small self-sustaining processor-based systems which collect information from their sensors, produce a computation set, and data relayed to a Base Station. The nodes deployment is done over a range of environment types extending from harsh to hostile. The network's requirements vary depending on the environment type. WSNs must have self-sufficiency and autonomy in harsh environments. Whilst, security is crucial in hostile environments, where the WSNs must be trustworthy and secure. In order to reduce production costs and decrease power usage, the design of nodes in WSNs is typically very simple. Sensor networks inherit all aspects of WSNs but also have their own unique features. Thus, the WSN security model design is quite distinctive from that of Ad hoc networks. In hostile environments, an Intrusion Detection System (IDS) is very vital for WSNs as it has the ability to identify malicious network packets. IDS can be efficiently employed in numerous methods like Neural Networks. Despite that, the classification algorithms must have the least cost of computation in resource-constrained environments. This work has proposed a novel clustering algorithm with an integrated IDS classifier using the modified Neural Network. The Neural Network structure can be optimized by the proposed System Mentoring-Learning-Based technique for detection of optimal cluster-heads, and enhancement of the intrusions' classification accuracy.