A Federated Learning-Based Intrusion Detection System Using Dynamic Ensemble Aggregation for IoT Networks
Marwa Baich, Nawal Sael · IEEE Access · 2025
Intrusion Detection Systems (IDS) are essential for securing Internet of Things (IoT) and Industrial IoT (IIoT) environments, where data privacy and scalability are major challenges for traditional centralized approaches. This study proposes a Federated Learning-based IDS (FL-IDS) that leverages a lightweight Convolutional Neural Network (CNN) tailored for tabular data to detect cyber-attacks while preserving data confidentiality. To address class imbalance at the client level, we apply SMOTE locally before training. The system integrates a dynamic ensemble aggregation strategy combining Boosting with Top-K client selection, allowing the global model to adaptively weigh client contributions based on their performance history. Experiments conducted on three benchmark datasets—NSL-KDD, Edge-IIoTset and CICIDS2017 — under varying federation scales reveal that our approach consistently achieves high performance, with accuracy exceeding 98% and strong precision, recall, F1-score, and Matthews Correlation Coefficient (MCC), while maintaining a minimal false alarm rate. The obtained results demonstrate the capability of dynamic ensemble aggregation in federated learning for developing privacy-preserving, scalable IDS solutions suitable for IoT and IIoT environments.