Edge-FLGuard: A Federated Learning Framework for Real-Time Anomaly Detection in 5G-Enabled IoT Ecosystems
Manuel J. C. S. Reis
Abstract
The rapid convergence of 5G networks and Internet of Things (IoT) technologies has unlocked unprecedented connectivity and responsiveness across smart environments—but has also amplified cybersecurity risks due to device heterogeneity, data privacy concerns, and distributed attack surfaces. To address these challenges, we propose Edge-FLGuard, a federated learning and edge AI-based anomaly detection framework tailored for real-time protection in 5G-enabled IoT ecosystems. The framework integrates lightweight deep learning models—specifically autoencoders and LSTM networks—for on-device inference, combined with a privacy-preserving federated training pipeline to enable scalable, decentralized threat detection without raw data sharing. We evaluate Edge-FLGuard using both public (CICIDS2017, TON_IoT) and synthetic datasets under diverse attack scenarios including spoofing, DDoS, and unauthorized access. Experimental results demonstrate high detection accuracy (F1-score ≥ 0.91, AUC-ROC up to 0.96), low inference latency (<20 ms), and robustness against data heterogeneity and adversarial conditions. By aligning edge intelligence with secure, collaborative learning, Edge-FLGuard offers a practical and scalable cybersecurity solution for next-generation IoT deployments.