Generative AI Sensor Fusion for Secure Digital Twin Ecosystems: A Standardization-Aligned Framework for Cyber-Physical Systems

Active fault management of the integration of Cyber-Physical Systems (CPSs), Digital Twins (DTs), and AI-driven sensors needs a secure and uniform framework. In relevance, the paper presents a hybrid Variational Autoencoder-Generative Adversarial Network (VAE-GAN)-based model with a redesigned Generative AI-Driven Sensor Fusion (GAI-SF) architecture. The proposed architecture combines the streams of multivariate sensor data and preempts anomalies before performance levels may be detected as degraded. Experimental validation based on Tennessee Eastman Process (TEP) and NASA turbofan engine data shows that it is up to 80 times better than classic fusion techniques in F1-score, and lowers the detection latency by some 40%. In addition to predictive performance, GAI-SF is fully compatible with ISO/IEC JTC 1/ SC 42, 3GPP SA6, and ETSI MEC standards, meaning it can be used in secure, interoperable, and AI-controlled deployment in next-generation 5G/6G CPS ecosystems.