An Efficient Intrusion Detection System using Dimensionality Reduction with Horse Optimization Based Ensemble Classifier for Internet of Things Devices

Cybersecurity has evolved progressively due to the proliferation of the Internet of Things (IoT), since numerous compact, intelligent gadgets transmit vast quantities of data to the Internet. Nevertheless, these systems have numerous defence faults that result from the absence of security tools and support for hardware security, making them vulnerable to cyberattacks. Hence, the growth of IoT devices to deliver security over resistance to threats is a requirement to create an IoT that is secure and effective. Defending these things is very significant for the security of the system. Moreover, it is significant to incorporate the Intrusion Detection System (IDS) with an IoT device. IDS aims to perceive and analyse network traffic from dissimilar sources and detect malicious actions. It is an important fragment of cybersecurity technology. Presently, the Deep Learning (DL)-based system plays a fundamental part in the IDS scheme to detect and classify attacks. This paper develops an Efficient Intrusion Detection System Using Dimensionality Reduction and an Ensemble-based Classification Model (EIDS-DRECM) for Internet of Things Devices. The main intention of the EIDS-DRECM paper is to enhance cybersecurity in IoT networks by developing efficient threat detection and mitigation mechanisms to ensure data integrity, confidentiality, and system resilience. The min-max scaler methodology has been employed initially in the data preprocessing stage by changing and organizing raw data into a suitable format. Followed by, the process of Feature Selection (FS) is executed by the Horse Optimization Algorithm (HOA) to retain the most relevant feature from a dataset. At last, the ensemble models of Graph Convolutional Network (GCN), Temporal Convolutional Network (TCN), and Deep Recurrent Q-Network (DRQN) are used for the attack detection process. An extensive set of simulations was involved in exhibiting the promising results of the EIDS-DRECM method. The experimentation results inferred the proficient performance of the EIDS-DRECM system in the attack detection procedure.