Anomaly Detection in Metaverse Healthcare for 6G-Enabled Internet of Things: An Approach Based on Feature Selection

The metaverse is based on emerging technologies and has immense potential to improve clinical and overall well-being outcomes. However, it is susceptible to attacks such as port scanning, probing, and distributed denial-of-service in health care analytics. These have to be addressed with efficient 6G-enabled intrusion detection systems that are capable of recognizing abnormal behavior in health care in the metaverse. The traditional anomaly detection approaches are impaired by class imbalance, poor feature selection, and hyperparameter tuning that limit their performance. In this work, a novel anomaly detection system with 6G technology is introduced that employs reinforcement learning (RL) to address class imbalance and attribute selection. The performance is optimized by utilizing the Bayesian Optimization Hyperband (BOHB) hyperparameter-tuning method. The process of RL is further improved by incorporating explain ability approaches for minority class boosting in classification and selection of significant features. The model is evaluated on MA WI and NSL-KDD data with improved accuracy, AVC and F-measure at 90.691 %, 92.423%, 0.817 and 90.604%, 93.565%, 0.822. These results confirm the model's ability to detect cyber threats in healthcare in the metaverse and pave the way for safer and more reliable digital health ecosystems.