Development of an effective intrusion detection system using hybrid scheme for wireless sensor networks

Abstract

One of the challenging research areas in the latest advancement in the newlineengineering field is Wireless Sensor Networks (WSN). WSN is used in a newlinewide range of applications such as traffic monitoring, battlefield, newlinesurveillance, environment monitoring, health care, etc. They are newlinecharacterized by constrained energy resources, communication, and newlinecomputational capabilities. They are vulnerable to attacks due to their newlinecharacteristics and deployment in a hostile and unattended environment. New newlineapproaches are being introduced to increase the overall security of the newlinenetwork. In general, Intrusion Prevention, Intrusion Detection, and Intrusion newlineMitigation secure WSNs. This research provides the algorithms and schemes newlinefor Intrusion Detection System (IDS) to detect intrusions in different levels newlineof the wireless sensor network. For security purposes, two approaches are newlineestablished for improving security. newlineIn the first approach, to protect the network against security threats, newlinean efficient IDS is designed. The NSL-KDD and UNSW-NB15 datasets newlineconsists of attributes about the entering and leaving packets of the network. newlineThe Stacked Auto Encoder processes the datasets to identify the promising newlineparameters to enhance the performance of IDS. Upon identifying the most newlineessential features, it is fed as input to the IDS model. In this method, RNN is newlinemodelled with an optimized feedback connected RNN model which serves newlineas the IDS system. The POA optimally selects the weight values to enhance newlinethe effectiveness of IDS. In the testing phase, the proposed model determines newlinewhether the input packets are intruded or not. newline