Hybrid of WOA-ABC and proposed CNN for intrusion detection system in wireless sensor networks

Abstract

Wireless sensor networks (i.e., WSNs) are self-configured and infrastructure-less wireless networks that monitor environmental or physical factors, such as sound, temperature, vibration, motion, pressure, or pollutants, and collaborate with the base station for analysis of their data. A sink, also known as a base station, serves as a link between customers and the network. However, WSNs have various challenges such as high energy consumption, high bandwidth demand, quality of service, compression techniques, data processing, and intrusion. To avoid intrusion, an intrusion detection system is the best solution. It investigates the network by gathering enough information from each node and detecting abnormal sensor node behavior. The intrusion detection system is primarily divided into two steps: feature extraction/selection and classification. In this work, we introduced a new intrusion detection system based on a deep learning network. For feature selection, a proposed hybrid WOA-ABC algorithm is used, and a proposed CNN architecture is used for classification. In WSN there are various attacks available mainly DoS attack, Sybil attack, black hole attack, wormhole attack, etc. In this paper, the NSLKDD dataset is used for intrusion detection. In the NSLKDD dataset, numerous attacks available but those attacks come under 4 types of attacks. There are Dos, UR2, R2L, and probe. So, in this work mainly we concentrate to classify these 4 types of attacks. Compared to the existing method our proposed architecture gives the best performance result in terms of execution time, detection rate, accuracy, and false alarm rate. When compared to existing approaches, our proposed mechanism minimizes execution time by 76.54 %.