MDR-TCN-BiGAtt model based IoT intrusion detection in fog environment

In recent years, fog computing and IoT convergence have contributed to increased cyber attacks. Thus, the fog nodes are vulnerable to high-dimensional anomalous traffic attacks due to their limited computing resources. To address this issue, we propose an intrusion detection model termed MDR-TCN-BiGAtt in this paper. Specifically, this model utilizes an improved feature selection algorithm for dimensionality reduction. It employs BiGRU to capture anterior-posterior temporal dependencies in features and includes a TCN layer to reduce computation and expand the data acceptance domain, enhancing the model's detection performance. Furthermore, we incorporate an attention mechanism into the model to determine the importance of assigned features, thereby enhancing the extraction of crucial features. To assess the performance of our model, we conducted evaluations using two widely accessible datasets, namely UNSWNB15 and IOT-23. The experimental results substantiate that our model outperforms existing models in accuracy, precision, and recall. Additionally, it exhibits faster convergence for intrusion detection in IoT within the fog computing environment.