Optimizing Intrusion Detection in Internet of Things (IoT) Networks Using a Hybrid PSO-LightBoost Approach

The IoMT in healthcare has greatly improved and enhanced patient monitoring and treatments.However, with this widespread adoption of the Internet of Medical Things (IoMT), the systems have introduced significant security issues, especially concerning intrusion detection systems in IoMT networks.The computational limitations of IoMT devices make it difficult to implement traditional security measures, which has led to the exploration of advanced machine learning and optimization techniques for better security.This research aims to enhance intrusion detection in IoMT environments by combining Particle Swarm Optimization (PSO) with LightGBM, known as LightBoost, into a PSO-LightBoost model.The main objective is towards improving detection accuracy along with scalability and efficiency in managing large-scale and complex data sets while minimizing false positives.Methodology: The suggested method has several data preprocessing steps involved.These include the use of Adaptive Synthetic Sampling Approach for Imbalanced Learning (ADASYN) for managing class imbalance and Z-score normalization for outlier detection purposes.Feature selection is done using a hybrid approach-Variance Mutual Information Random Forest (VarMiRF)-for establishing the most relevant features.Further optimization of the LightGBM model's hyperparameters was done by the PSO algorithm to enhance its performance.It reveals that PSO-LightBoost produces great performance metrics, achieving an accuracy of 97.28%, precision of 93.49%, and recall of 91.88% on the UNSW-NB15 dataset, a comprehensive dataset designed for network intrusion detection.Based on the findings, it was concluded that the proposed hybrid model of PSO-LightBoost is a reliable solution to intrusion detection within an IoMT environment, showing higher accuracy and efficiency in computing.The future work concerns the extension of this model towards real-time attack detection and adversarial classification.