Industrial deployments increasingly rely on Open Platform Communications Unified Architecture (OPC UA) as a secure and platform-independent communication protocol, while private Fifth Generation (5G) networks provide low-latency and high-reliability connectivity for modern automation systems. However, their combination introduces new attack surfaces and traffic characteristics that remain insufficiently understood, particularly with respect to machine learning-based intrusion detection systems (ML-based IDS). This paper presents an experimental study on detecting cyberattacks against OPC UA applications operating over an operational private 5G network. Multiple attack scenarios are executed, and OPC UA traffic is captured and enriched with statistical flow-, packet-, and protocol-aware features. Several supervised ML models are trained and evaluated to distinguish benign and malicious traffic. The results demonstrate that the proposed ML-based IDS achieves high detection performance for a representative set of OPC UA-specific attack scenarios over an operational private 5G network.
Machine learning (ML) based anomaly detection has been thoroughly studied in the literature where different architectures and models have been developed for different industrial cyber-physical systems (CPS). However, given that most CPS control systems are deployed on embedded platforms with limited storage and computational resources, deployment and real-time functionality remain challenging. This paper applies a combination of TinyML techniques, Pruning, and quantization to reduce the computation cost and memory footprint of a transformer-based anomaly detection for an industrial combustion system without degrading the accuracy and classification performance. To assess reliability, we compare the reduced model's performance and accuracy to the original model and the state-of-the-art Cumulative Sum Control Chart (CUSUM) analytical model. The experiment results demonstrate that, while reducing the memory footprint by 70% and computation cost by 8%, the resulting model does not degrade the classification accuracy. Moreover, the reduced transformer model outperforms the CUSUM classification accuracy by up to 99%.
Today's manufacturing environments are challenged by the adoption of upcoming data-driven technologies, such as artificial intelligen0ce or digital twins. The bottleneck for digital transformation are mostly traditionally grew industrial hardware and software systems. The current pace of changes in technologies, requires a strong adaptive architecture backbone to address rapidly changing tools for manufacturing. This paper proposes a cloud native architecture for manufacturing, using Kubernetes as orchestration platform for industrial applications. Since strict requirements need to be met for control and monitoring, the proposed approach includes a layer model for multi-cluster deployments able to fulfill divers requirements. A holistic multicluster controller is presented, which distributes applications on the respective layers, based on their needs.
The ADACOR holonic architecture introduced an adaptive reconfigurable manufacturing control approach that promotes a balance between centralized and decentralized control, reducing the reliance on a pure central node and enhancing the robustness, responsiveness and fault tolerance. In the digital transformation era, it is crucial to understand the alignment of ADACOR with Industry 4.0 and Industry 5.0 principles. In this context, the paper discusses the alignment of ADACOR with the RAMI4.0 model, as well as with the human-centric and sustainability dimensions defined by Industry 5.0. Finally, the research challenges to develop an ADACOR 4.0 architecture fully aligned with these principles were also discussed.
Industrial Cyber-Physical Systems increasingly rely on Machine Learning models-particularly generative approaches such as Variational Autoencoders - for critical tasks such as anomaly detection and predictive maintenance. However, these models are vulnerable to adversarial attacks that can covertly mislead classification, posing severe risks in safetycritical domains. This paper introduces a novel Adversarial Sample Generation Algorithm (ASGA), specifically designed to craft adversarial samples for Variational Autoencoder-based anomaly detection systems used in industrial time-series applications. ASGA uses an optimization process that iteratively adjusts abnormal samples by applying small, targeted perturbations to minimize reconstruction errors, ensuring that anomalies are misclassified as nominal system functions. The proposed algorithm incorporates techniques such as gradient-based optimization, perturbation projection, and adaptive hyperparameter tuning to maintain stealthiness and stability. We evaluate ASGA on two industrial datasets, demonstrating consistent reductions in key anomaly detection metrics (accuracy, F1-score, ROC-AUC) and high attack success rate of 70 %. These results highlight the efficacy of ASGA and emphasize the urgent need for resilient anomaly detection mechanisms to safeguard ICPS against sophisticated adversarial threats.
