Metaheuristic-Enhanced XGBoost Framework for Intrusion Detection in Smart Home IoT Systems

Abstract

The growing integration of smart technologies into residential environments has heightened their exposure to cybersecurity threats, thereby necessitating robust and intelligent intrusion detection systems (IDS). This study proposes a hybrid AI-driven intrusion detection framework tailored for smart home networks, leveraging the Extreme Gradient Boosting Classifier (XGBoost or XGBC) enhanced by three metaheuristic optimization techniques: the Arithmetic Optimization Algorithm (AOA), Horse Herd Optimization (HHO), and Wild Geese Algorithm (WGA). The performance of these hybrid schemes XGAO, XGHH, and XGWG was rigorously evaluated using a comprehensive dataset containing 148,518 labeled network traffic instances, compiled through data mining methods. The dataset includes diverse attributes such as source and destination bytes, service types, protocols, and various connection-related flags. Performance evaluation was conducted using four standard classification metrics: accuracy, precision, recall, and F1-score. Among all models, the hybrid XGAO scheme demonstrated superior performance, achieving a training accuracy of 0.991, outperforming the baseline XGBC model, which scored 0.946 under the same conditions. In the testing phase, XGAO maintained high generalizability with an accuracy of 0.987, compared to 0.953 for XGBC. The XGAO model also excelled in recall (0.989), precision (0.991), and F1-score (0.990) for correct detections. These findings affirm that integrating XGBoost with AOA significantly enhances the classification accuracy and reliability of intrusion detection systems in smart home environments. The study contributes to the development of resilient, adaptive IDS architectures capable of mitigating evolving cyber threats in the domain of intelligent residential technologies.