Enhancing Industrial-IoT Cybersecurity Through Generative Models and Convolutional Neural Networks

The Industrial Internet of Things (I-IoT) has revolutionized manufacturing, enhancing efficiency and productivity. However, this transformation has introduced a critical challenge: securing the vast interconnectedness of industrial systems. Traditional security measures struggle in the complexities of I-IoT environments, leading to the exploration of machine learning and deep learning models trained on intrusion detection system logs. These intelligent approaches autonomously detect and respond to threats, adapting to evolving attack patterns. Unfortunately, the robustness of models in the field of I-IoT depends not only on their detection accuracy but also on their requirement to be both non-complex and lightweight. Furthermore, the quality and diversity of training datasets heavily influence the robustness of these models. In this paper, we focus on these three main challenges. Recognizing the characteristics of I-IoT environments, we propose a lightweight, non-complex CNN1D model. Our model has been trained on the diverse TII-SSRC-23 Dataset, which we balanced using the Conditional Tabular GAN to strengthen the model’s generalization capabilities. Our approach yields outstanding results, achieving a remarkable 100% accuracy, precision and F1-score rates in addressing the security challenges posed by the Industrial Internet of Things. Furthermore, the performance of our model was validated using a cross-validation technique.