A Comparative Analysis of Ensemble Learning Techniques for Enhanced Anomaly Detection in IoT Systems

Abstract

With the rapid proliferation of Internet of Things (IoT) devices, ensuring their security against potential anomalies has become paramount. This study employs a rigorous empirical approach to compare various ensemble learning techniques—specifically Random Forest, Gradient Boosting, and Stacking—focused on their efficacy in treating anomaly detection challenges within IoT frameworks. The dataset utilized comprises a rich array of recorded IoT network traffic generated under different attack scenarios. Results indicate that Stacking outperforms other methods with an accuracy rate exceeding 95%, while Random Forest demonstrates commendable latency metrics, achieving detection within milliseconds. These findings underscore the necessity for advanced anomaly detection mechanisms that integrate machine learning, thereby enhancing the security posture of IoT environments. The implications extend to both industrial applications and ongoing research in the field of cybersecurity.