An intellectual intrusion detection system using Hybrid Hunger Games Search and Remora Optimization Algorithm for IoT wireless networks

摘要

Smart devices are connected to the internet, called the Internet of Things (IoT); it is a wireless network that plays a vital role in modernizing people’s interaction by improving agility, mobility, efficiency and effectiveness. Although IoT offers various advantages, it also raises privacy and security problems in IoT networks. The security of such integrated networks still remains an active area of research. Information transported across a wireless network must be monitored and implemented efficiently to ensure its integrity, validity and confidentiality. In order to minimize the rate of security problems in IoT networks, an intrusion detection model is suggested. For that reason, various intrusion detection based techniques are introduced in existing works to improve the security of IoT networks. But existing methods do not provide an efficient result in securing IoT data. The main contribution of using intrusion detection techniques in the proposed work is to improve the security of IoT networks. In this work, a hybrid Hunger Games Search and Remora Optimization Algorithm (HHGS-ROA) is proposed to address the security problems. The proposed approach uses a real-world Aegean Wi-Fi Intrusion Dataset (AWID) for an IoT intruder detection system. Initially, the AWID raw captured traffic is fed into the pre-processing stage to increase the speed of feature selection. In the proposed approach, a hybrid HGS-ROA model is used to extract the features from the AWID dataset. Finally, the SVM classifier is proposed to classify network traffic as normal or malicious attacks. Therefore proposed approach achieves a low false-positive rate (FPR) with an enhanced detection rate. The proposed work is implemented in the PYTHON platform. Compared with existing methods in terms of accuracy, F1 score, detection rate, FAR, computation times, false alarm rate, MCC, and precision has proven to be more effective. The proposed approach performs better than existing algorithms in terms of evaluating metrics such as accuracy (99.16%), precision (99.76%), recall (99.40%), F1 score (99.58%), false alarm rate (0.00005%), MCC (99.97%) and false-positive rate (0.20%).