Adaptive Gaussian One-Class Classifier for Detecting Counterfeit Digital Twins

Abstract

Digital Twins (DTs) are highly realistic digital replications of physical objects, environments, and devices that can enhance our current experience level through the industrial metaverse (IMV). In addition to their applications for real-time analysis and monitoring, DTs are also helpful in risk prevention and securing identity. However, current DT-based security solutions suffer from scalability and deployment challenges. This paper leverages the unique behavioural properties of DTs to address such limitations through a one-class classification (OCC) system based on the Gaussian probabilistic model. The proposed OCC approach offers an adaptive and robust DT verification by identifying fake DT clone instances based on unusual behavioural patterns. Experimental results show that with an accuracy of 95% and a relatively higher completeness score, the proposed OCC method also outperforms the existing SVDD classifier. This work plays a pivotal role in shaping secure solutions using DTs and opens new dimensions of research for metaverse security.