RUDRA—A Novel Re-concurrent Unified Classifier for the Detection of Different Attacks in Wireless Sensor Networks

Abstract

Wireless Sensor networks find its applications in most fields such as health care, automotive, consumer electronics and most importantly Industry 4.0 automations. Nowadays integration of Internet of things (IoT), Wireless sensor networks (WSN) has become the most prominent and ubiquitous in day to day life, but these systems still suffer from the different security attacks, which makes the connected devices under immense pressure for an efficient and secured data transfer. To overcome this issue, intrusion detection system using hybrid artificial intelligence algorithm has been proposed for the better detection and classification. The paper proposes the most intelligent attack detection system (IADS) RUDRA which works on the principle of recon current LSTM networks (Long Short-Term memory) along with extreme learning machines (ELM) which is then used for detection of the different DoS attacks such as Sybil, wormhole, black hole, sinkhole and selective forwarding attacks. The proposed tool works on three different phases, such as feature decomposition, hybrid learning and decision phase. The real-time datasets were collected on the test bed which consists of RISC architecture as main CPU interfaced with CC2540 transceivers. Also, the proposed tool integrated with the hybrid classifier has been compared with other existing algorithms such as RNN-LSTM, ELM and SVM in which the accuracy of 98.4% is obtained for the proposed classifier.