Trust Beyond Perfection: Managing Uncertainty in the Internet of Digital Twins

Digital Twins (DTs) are increasingly adopted in IoT systems as autonomous agents capable of providing services and interacting within distributed ecosystems. However, most trust management solutions rely on overly idealized assumptions: service evaluations are always accurate, and benevolent nodes never make mistakes. These assumptions limit the applicability of such models in realistic scenarios, where uncertainty and imperfection are the norm. This paper proposes a trust management model tailored to more realistic DTs, where both service providers and requesters may introduce errors. Providers may fail due to physical or software limitations; requesters cannot rely on objective ground truth to evaluate outcomes. To capture this complexity, we introduce a feedback mechanism based on divergence from aggregated results, rather than binary correctness. We define an IoT scenario where DTs dynamically offer and request services. A trust computation model is applied to distinguish truly malicious nodes from those affected by faults, reducing false positives. Simulations confirm improved trust accuracy and system resilience under realistic conditions.