Hybrid Deep Learning Framework for Intelligent Intrusion Detection and Threat Mitigation in IoT Networks

The rapid growth and heterogeneity of Internet of Things (IoT) networks have increased the attack surface of modern networks and their exposure to diverse and evolving cyber threats. To cope with this change, the need for intrusion detection systems capable of identifying both known and previously unseen attacks has emerged. This paper proposes AEFS-Stack IDS, an advanced hybrid deep learning and machine learning framework for intrusion detection and threat mitigation in IoT networks. The proposed framework combines Deep Learning (DL) and Machine Learning (ML) architectures by integrating feature selection, autoencoder-based anomaly detection, and stacked classification to reduce the feature space, identify anomalous traffic, and classify known attack categories. In addition to this, a mitigation module is further incorporated to respond to normal, known malicious, and anomalous traffic based on the detection outcome. The framework is evaluated using the UNSW-NB15 and ToN-IoT benchmark datasets, with performance assessed using accuracy, precision, recall, F1-score, and computational measures. Experimental results show that AEFS-Stack IDS achieves 97.43% accuracy, 100% precision, and an F1-score of 98.70% on UNSW-NB15, while RF-CNN-LSTM achieves 99.90% accuracy, 100% precision, and an F1-score of 100% on ToN-IoT. The results demonstrate that integrating feature reduction, anomaly detection, heterogeneous classification, and threat mitigation within a unified framework are effective for real-time intrusion detection in resource-constrained IoT environments.