This paper provides the complete details of current challenges and solutions in the cybersecurity of cyber-physical systems (CPS) within the context of the IIoT and its integration with edge computing (IIoT–edge computing). We systematically collected and analyzed the relevant literature from the past five years, applying a rigorous methodology to identify key sources. Our study highlights the prevalent IIoT layer attacks, common intrusion methods, and critical threats facing IIoT–edge computing environments. Additionally, we examine various types of cyberattacks targeting CPS, outlining their significant impact on industrial operations. A detailed taxonomy of primary security mechanisms for CPS within IIoT–edge computing is developed, followed by a comparative analysis of our approach against existing research. The findings underscore the widespread vulnerabilities across the IIoT architecture, particularly in relation to DoS, ransomware, malware, and MITM attacks. The review emphasizes the integration of advanced security technologies, including machine learning (ML), federated learning (FL), blockchain, blockchain–ML, deep learning (DL), encryption, cryptography, IT/OT convergence, and digital twins, as essential for enhancing the security and real-time data protection of CPS in IIoT–edge computing. Finally, the paper outlines potential future research directions aimed at advancing cybersecurity in this rapidly evolving domain.

Industrial Internet of things (IIoT) environments need to implement reliable security measures because of the growth in network traffic and overall connectivity. Accordingly, this work provides the architecture of network traffic analysis and the detection of intrusions in a network with the help of edge computing and using machine-learning methods. The study uses k-means and DBSCAN techniques to examine the flow of traffic in a network and to discover several groups of behavior and possible anomalies. An assessment of the two clustering methods shows that K-means achieves a silhouette score of 0.612, while DBSCAN achieves 0.473. For intrusion detection, k-nearest neighbors (KNN), random forest (RF), and logistic regression (LR) were used and evaluated. The analysis revealed that both KNN and RF yielded seamless results in terms of precision, recall, and F1 score, close to the maximum possible value of 1.00, as demonstrated by both ROC and precision–recall curves. Accuracy matrices show that RF had better precision and recall for both benign and attacks, while KNN and LR had good detection with slight fluctuations. With the integration of edge computing, the framework is improved by real-time data processing, which means a lower latency of the security system. This work enriches the knowledge of the IIOT by offering a detailed solution to the issue of cybersecurity in IoT systems, based on well-grounded performance assessments and the right implementation of current technologies. The results thus support the effectiveness of the proposed framework to improve security and provide tangible improvements over current approaches by identifying potential threats within a network.

The Industrial Internet of Things (IIoT), combined with the Cyber-Physical Systems (CPS), is transforming industrial automation but also poses great cybersecurity threats because of the complexity and connectivity of the systems. There is a lack of explainability, challenges with imbalanced attack classes, and limited consideration of practical edge–cloud deployment strategies in prior works. In the proposed study, we suggest an Impact-Aware Taxonomy-Driven Machine Learning Framework with Edge Deployment and SHapley Additive exPlanations (SHAP)-based Explainable AI (XAI) to attack detection and classification in IIoT-CPS settings. It includes not only unsupervised clustering (K-Means and DBSCAN) to extract latent traffic patterns but also supervised classification based on taxonomy to classify 33 different kinds of attacks into seven high-level categories: Flood Attacks, Botnet/Mirai, Reconnaissance, Spoofing/Man-In-The-Middle (MITM), Injection Attacks, Backdoors/Exploits, and Benign. The three machine learning algorithms, Random Forest, XGBoost, and Multi-Layer Perceptron (MLP), were trained on a real-world dataset of more than 1 million network traffic records, with overall accuracy of 99.4% (RF), 99.5% (XGBoost), and 99.1% (MLP). Rare types of attacks, such as injection attacks and backdoors, were examined even in the case of extreme imbalance between the classes. SHAP-based XAI was performed on every model to help gain transparency and trust in the model and identify important features that drive the classification decisions, such as inter-arrival time, TCP flags, and protocol type. A workable edge-computing implementation strategy is proposed, whereby lightweight computing is performed at the edge devices and heavy, computation-intensive analytics is performed at the cloud. This framework is highly accurate, interpretable, and has real-time application, hence a robust and scalable solution to securing IIoT-CPS infrastructure against dynamic cyber-attacks.

Industrial Internet of Things (IIoT) networks have greatly expanded with emerging environments, including Internet of Vehicles (IoV), and the need to ensure real-time cybersecurity has become much more complex. We propose a robust and explainable cybersecurity framework for IoV in IIoT Cyber-Physical Systems (CPS), in which we perform implementation and validation using the RT-IoT2022 (Real-Time Internet of Things) dataset. The framework includes clustering techniques for pattern discovery, predictive modeling for threat detection, Explainable Artificial Intelligence (XAI) for interpretability enhancement, and a cloud–edge integration concept for efficient and scalable processing. The behavior patterns in the network traffic were discovered to be clustered, which helped in early anomalous behavior detection. Gaussian Mixture Modeling (GMM) achieved a Silhouette Score of 0.53 and an ARI of 0.75, indicating better alignment with the true labels. Decision Tree (DT), K-Nearest Neighbors (KNN), and Light Gradient Boosting Machine (LightGBM), as well as Multi-Layer Perceptron (MLP) classifiers, were used to carry out predictive analysis. Among these, DT attained the highest classification accuracy of 99.43%, followed by KNN (99.25%), MLP (98.90%), and LightGBM (81.03%). The DT, KNN, and MLP models achieved consistently high precision, recall, and F1-scores across most attack classes, whereas LightGBM exhibited comparatively lower performance for several classes. In order to maintain transparency in decision-making, LIME (Local Interpretable Model-agnostic Explanations) was used to get feature-level insights into the predictions of each model. The LIME analysis shows that each model relied on several features and decision-making logic, such as threshold-based splits in DT, feature similarity in KNN, volume-based patterns in LightGBM, and non-linear interactions in MLP. The proposed framework combines cloud–edge processing, high detection accuracy, early threat identification, and explainability, and thus is a powerful solution for real-time IIoT cybersecurity applications.

A key challenge in Internet of Vehicles (IoV) environments is accurately predicting collision risks, given the dynamic nature of traffic systems and the variety of available data sources. Existing approaches often rely on structured IoV data, while visual contextual information is not fully incorporated into the prediction process. This limitation motivates a unified framework that integrates environmental perception with sensor-derived information to enable more accurate collision-risk prediction. In this study, we propose a two-stage context-aware meta-ensemble framework that combines image-based perception and IoV-based predictive analysis for collision-risk prediction. The first stage uses multiple pre-trained convolutional neural networks to obtain rich visual representations from vehicle images through feature-level fusion. The second stage employs a hybrid stacked ensemble architecture that combines the probabilistic outputs of heterogeneous machine learning models with IoV and image-derived features, which a meta-learner then processes for final collision-risk prediction. The proposed feature-fusion model performs best in the perception stage, with an accuracy of 0.97. In the IoV-based prediction stage, the proposed stacked ensemble architecture outperforms all baseline models, achieving an accuracy of 0.9847. Robustness analysis shows that the perception module tolerates blur and rotation but remains sensitive to Gaussian noise. Cross-stage integration incorporates image-derived features into the IoV feature space, providing complementary visual context for collision-risk prediction and demonstrating its potential for intelligent transportation and cyber-physical IoV applications.