Cognitive neural synthesis for Industrial Internet of Things security control and intrusion detection
Abdullah Alwabli · Journal of Engineering Research · 2026
High-density Industrial Internet of Things (IIoT) devices have increased security risks, especially regarding risks of intrusions and protection against threats in real time. Static approaches cannot address the scale issues, the imbalance in aspects of the IIoT traffic, and the inherent intricate structure that is impossible for conventional algorithms to decipher independently and is better suited for modern security requirements. In this paper, we develop a novel architecture known as the Cognitive Neural Synthesis (CNS) for improving intrusion detection in IIoT systems. Deep learning procedures involve Transformer models for self-attention, Gated Recurrent Unit (GRU) for temporal dependency, and DenseNet, solving the weaknesses of classic and isolated deep learning models. In this study, the performance of the proposed CNS Framework was assessed using three standard datasets, such as NSL-KDD, CICIDS2017, and TON_IoT, to generalize the proposed framework for different IIoT scenarios. The data collected was normalized, and while both the oversampling and under-sampling techniques were implemented in a hybrid manner to solve the class imbalance problem, the feature extraction used included Mutual Information (MI) and Gradient Boosting (GB). The performance of the model was benchmarked against the previous approaches, such as the convolutional neural networks (CNNs), recurrent neural networks (RNNs), and the traditional classifiers, such as the support vector machine (SVM) and Decision Tree (DT), in terms of accuracy, recall, F1-score, and the area under the curve (AUC). The proposed CNS Framework was significantly more accurate in training (99.1) and validation (98.5) on NSL-KDD. On the CICIDS2017 dataset, the proposed framework achieved strong ROC-AUC performance, demonstrating robust intrusion detection capability. Also, this framework had low training time and scalability, enabling its application to real-world IIoT systems. The CNS Framework is sufficient to address securitization motives associated with IIoT threats and vulnerabilities by providing CNS architectural elements that enable a highly competent, adaptable, and efficient ID mechanism. The results presented herein suggest the potential of advanced hybrid deep learning architectures to enhance the resilience and reliability of IIoT networks.