IoT Network Security: Ensemble-Based Approaches for Anomaly Detection
Lamya Jaafouri, Maryam Et-Tolba, Charifa Hanin, Iyad Lahsen-Cherif · 2025
IoT networks have transformed various sectors by interconnecting billions of devices, yet they remain highly suscep-tible to cyber threats such as Distributed Denial of Service (DDoS) attacks, ransomware, and phishing. These vulnerabilities compro-mise the integrity, confidentiality, and availability of IoT systems, necessitating scalable defense mechanisms. This paper presents an ensemble learning-based framework for anomaly detection in IoT networks, optimizing both accuracy and computational effi-ciency for real-time security applications. The proposed approach integrates Gradient Boosting, Random Forest, and Isolation Forest to capture complex patterns in network traffic. Experi-mental validation on the IoT-23, CICIDS2017, and UNSW-NB15 datasets demonstrates its superior performance over state-of-the-art models. Notably, on the UNSW-NB15 dataset, the framework achieves an accuracy of 88.37%, an Fl-score of 88%, and an AUC of 95.6 %, surpassing existing approaches. These results highlight ensemble learning's potential to strengthen IoT cybersecurity by leveraging multiple models for a more resilient defense against sophisticated cyber threats. The findings emphasize the necessity of integrating advanced machine learning techniques into IoT security frameworks to address the continuously evolving cyber threat landscape.