HFEL: A hybrid federated ensemble learning framework for intrusion detection in IoT networks

HFEL: A hybrid federated ensemble learning framework for intrusion detection in IoT networks

연구 분야: Cryptography

학회: Cluster Computing

Recent advancements in Internet of Things (IoT) networks have introduced significant security challenges, particularly in detecting and preventing cyber-attacks while preserving data privacy. This paper presents a novel Hybrid Federated Ensemble Learning (HFEL) framework that combines the privacy-preserving benefits of federated learning with the robust detection capabilities of ensemble methods. Our approach integrates three complementary models— Multi-Layer Perceptron, Random Forest, and XGBoost—in a federated learning environment, utilizing majority voting for final decision-making. The framework is extensively evaluated on two benchmark datasets: BoT-IoT and Edge-IIoT. Our experimental results demonstrate that HFEL significantly outperforms traditional federated learning approaches, achieving 99.94% accuracy on the BoT-IoT dataset and 97.53% on the Edge-IIoT dataset. Notably, HFEL maintains consistent performance from initial deployment, addressing the cold-start problem common in federated learning systems. Comparative analysis with state-of-the-art methods confirms HFEL’s superior detection capabilities while preserving data privacy, making it particularly suitable for securing distributed IoT networks.