Metaverse-IDS: Deep learning-based intrusion detection system for Metaverse-IoT networks

Abstract

Combining the metaverse and the Internet of Things (IoT) will lead to the development of diverse, virtual, and more advanced networks in the future. The integration of IoT networks with the metaverse will enable more meaningful connections between the 'real' and 'virtual' worlds, allowing for real-time data analysis, access, and processing. However, these metaverse-IoT networks will face numerous security and privacy threats. Intrusion Detection Systems (IDS) offer an effective means of early detection for such attacks. Nevertheless, the metaverse generates substantial volumes of data due to its interactive nature and the multitude of user interactions within virtual environments, posing a computational challenge for building an intrusion detection system. To address this challenge, this paper introduces an innovative intrusion detection system model based on deep learning. This model aims to detect most attacks targeting metaverse-IoT communications and combines two techniques: KPCA (Kernel Principal Component Analysis which was used for attack feature extraction and CNN (Convolutional Neural Networks for attack recognition and classification. The efficiency of this proposed IDS model is assessed using two widely recognized benchmark datasets, BoT-IoT and ToN-IoT, which contain various IoT attacks potentially targeting IoT communications. Experimental results confirmed the effectiveness of the proposed IDS model in identifying 12 classes of attacks relevant to metaverse-IoT, achieving a remarkable accuracy of 99.8% and a False Negative Rate FNR less than 0.2. Furthermore, when compared with other models in the literature, our IDS model demonstrates superior performance in attack detection accuracy.