Resource-Constrained Secure Graph Neural Clustering for Industrial Manufacturing

Abstract

Industrial manufacturing systems increasingly rely on graph-structured data derived from machines, sensors, and operational technology (OT) networks to support monitoring, optimisation, and anomaly analysis. However, deploying graph neural network (GNN)–based clustering methods in such environments is challenging due to strict resource constraints on edge and control hardware, as well as heightened security risks arising from compromised or noisy nodes. Existing graph clustering approaches typically assume abundant computational resources and benign data conditions, limiting their applicability in real-world industrial settings. In this work, we propose a resource-constrained secure graph neural clustering framework tailored for industrial manufacturing systems. The proposed method integrates lightweight graph neural representations with security-aware constraints that mitigate the influence of adversarial perturbations, faulty devices, and unreliable communication links. By explicitly accounting for memory, computation, and latency limitations, the framework enables stable and efficient clustering on OT-grade hardware without sacrificing robustness. Extensive experiments on industrial-style graph datasets demonstrate that the proposed approach achieves competitive clustering quality while significantly improving resilience under resource scarcity and security stress. The results highlight the practicality of secure GNN-based clustering for deployment in real manufacturing environments, bridging the gap between advanced graph learning techniques and operationally constrained industrial systems.