Abstract
The seamless interaction between the virtual and real worlds is due to the unprecedented degrees of decentralization, immersiveness and connectedness made possible by the Internet of Things (IoT) and the metaverse. In this light, it brings important ethical, privacy, and security considerations into play, hence calling for the strong protection of IoT-enabled metaverse systems. Anomaly detection is critical for solving the aforementioned issues and ensuring the dependability and security of the connected devices by identification and preventing malicious activity in IoT networks. With IoT networks being highly dynamic and complex, robust anomaly detection frameworks are essential for ensuring security and trust in the metaverse. This paper proposed a hybrid model combining Random Forest (RF) and Neural Network (NN) and compared it with a variety of machine learning (ML) techniques including Decision Tree (DT), Naive Bayes (NB), K-Nearest Neighbor (KNN), RF and Logistic Regression (LR) to detect anomalies in IoT-enabled metaverse environments. These models were trained and tested using the CIC-IDS 2017 Network Intrusion Dataset, a comprehensive benchmark used for evaluating intrusion detection systems (IDS). Indeed, with outstanding accuracy equaling a staggering 99.99%, the proposed hybrid model algorithm performed better than other ML models under study. This illustrates its vast potential for high-accuracy anomaly identification and false positives.