Improving Fog Computing Security with Deep Learning
Ali-Alridha Khalil, Mehdi Ebady Manaa · Engineering Technology & Applied Science Research · 2025
The rapid growth of the Internet of Things (IoT) has introduced new security challenges in distributed and resource-limited environments, most notably at the fog layer. Moreover, traditional Intrusion Detection Systems (IDS), which typically rely on cloud-based architectures and signature-based detection, are inadequate for meeting the latency, bandwidth, and adaptability requirements of Industrial IoT (IIoT) systems. In this research, we propose a hybrid Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) model tailored for fog-layer deployment. The model employs CNNs to extract local spatial features from network traffic and LSTMs to capture temporal dependencies associated with evolving threats. Evaluation is conducted using the Edge-IIoTset dataset, a comprehensive benchmark containing realistic IIoT traffic and 15 diverse attack types. Through extensive preprocessing, Chi-Squared (χ2)-based feature selection, and architectural fine-tuning, the model achieves 100% accuracy, precision, recall, and F1-score in binary classification, achieving high-fidelity detection with low false positives and minimal computational overhead. These results validate the proposed model as a robust and scalable security mechanism for fog-based IIoT environments.