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 engineering field is Wireless Sensor Networks (WSN). WSN is used in a wide range of applications such as traffic monitoring, battlefield, surveillance, environment monitoring, health care, etc. They are characterized by constrained energy resources, communication, and computational capabilities. They are vulnerable to attacks due to their characteristics and deployment in a hostile and unattended environment. New approaches are being introduced to increase the overall security of the network. In general, Intrusion Prevention, Intrusion Detection, and Intrusion Mitigation secure WSNs. This research provides the algorithms and schemes for Intrusion Detection System (IDS) to detect intrusions in different levels of the wireless sensor network. For security purposes, two approaches are established for improving security. newlineIn the first approach, to protect the network against security threats, an efficient IDS is designed. The NSL-KDD and UNSW-NB15 datasets consists of attributes about the entering and leaving packets of the network. The Stacked Auto Encoder processes the datasets to identify the promising parameters to enhance the performance of IDS. Upon identifying the most essential features, it is fed as input to the IDS model. In this method, RNN is modelled with an optimized feedback connected RNN model which serves as the IDS system. The POA optimally selects the weight values to enhance the effectiveness of IDS. In the testing phase, the proposed model determines whether the input packets are intruded or not. newline