A hybrid temporal deep learning framework using TCN–GRU and attention for intrusion detection in IoT networks
DOI:
https://doi.org/10.18488/76.v13i4.5372Keywords:
Anomaly detection, Cybersecurity, Deep learning, Internet of things, Intrusion detection system, Lightweight security models, Network traffic analysis.Abstract
The swift expansion of Internet of Things applications has significantly increased the number and complexity of cyberattacks, creating a pressing need for reliable and efficient intrusion detection systems. This study aims to develop a lightweight hybrid deep learning framework that improves intrusion detection. The proposed model combines TCN, GRU, and an attention mechanism to learn important temporal features from network traffic. To address class imbalance in the BoT-IoT dataset, SMOTE is applied during training to improve the representation of minority attack classes. The proposed framework is evaluated using standard classification metrics, including accuracy, precision, recall, F1-score, and computational efficiency. Experimental results demonstrate that the model achieves 99.92% accuracy, along with high precision and F1-score, while requiring only 15K trainable parameters, making it computationally efficient. The findings indicate that combining temporal feature extraction with attention-based feature weighting improves the detection of diverse IoT attacks without increasing model complexity. The proposed framework offers a practical and scalable solution for deployment in resource-constrained IoT environments, where accurate and efficient real-time intrusion detection is essential.
