LightEnsemble-Guard: An Optimized Ensemble Learning Framework for Securing Resource-Constrained IoT Systems

The rapid proliferation of real-time Internet of Things (IoT) devices has increased the need for efficient and accurate security mechanisms. Real-time IoT devices are highly vulnerable to cyberattacks due to their continuous connectivity and limited security mechanisms. In this work, we propose LightEnsemble-Guard, an ensemble learning-based approach specifically designed for resource-constrained IoT environments. Our method achieves a high detection accuracy of 99.55%, with strong precision, recall, and F1-score, while maintaining a low computational cost of just 22.23 seconds. To ensure robustness, we conducted 5-fold cross-validation and ROC curve evaluations, confirming the model’s reliability and generalizability. LightEnsemble-Guard integrates three lightweight classifiers LightGBM, XGBoost, and Extra Trees using a majority voting mechanism to improve detection performance on highly imbalanced datasets without burdening system resources. This ensemble strategy ensures an optimal balance between detection performance and computational resource utilization, making it well-suited for IoT networks, where processing power and memory are limited. The results highlight LightEnsemble-Guard as an effective, scalable and lightweight solution for real-time IoT security, significantly outperforming traditional models in both accuracy and computational efficiency.