Selfish Herd Optimization with Improved Deep Learning based Intrusion Detection for Secure Wireless Sensor Network
Suma Christal Mary S, E. Thenmozhi, K Murugeswari, N. Senthamilarasi · International Journal of Electronics and Communication Engineering · 2023
Wireless sensor networks (WSNs) are becoming more frequently utilized in many applications like environmental monitoring, smart cities, and healthcare. But, security is an important problem in WSNs because of the possible vulnerability to attacks. Intrusion detection systems (IDS) are utilized for detecting and preventing attacks on WSNs. Typical IDS depend on rule-based or signature-based methods that are limited in their capability for detecting before unseen attacks. Deep learning (DL)-based IDS are exposed to promising outcomes in identifying novel attacks. DL-based IDS for WSNs are created utilizing an integration of supervised and unsupervised learning approaches. Therefore, this study designs a Selfish herd optimization with Improved Deep Learning based Intrusion Detection (SHOIDL-ID) technique for secure WSN. The presented SHOIDL-ID technique focuses on the process of identifying and classifying intrusions in the WSN. The presented SHOIDL-ID approach applies data preprocessing to normalize the input data to accomplish this. The SHOIDL-ID technique employs an attention-based bidirectional long short-term memory (ABiLSTM) approach for intrusion recognition and classification. Finally, the SHO approach was utilized for the optimal hyperparameter tuning of the ABiLSTM algorithm. The experimental validation of the SHOIDL-ID approach takes place on the WSN-DS dataset. The outcomes indicate the improved performance of the SHOIDL-ID methodology over other existing approaches in terms of different measures.