Scalable Dynamic Trust Evolution Framework for Large-Scale IoT Information Systems with Asymmetric Evidence Weighting

INTRODUCTION: The growth of Internet-of-Things (IoT) infrastructure being built pushes the envelope to giant scale, heterogeneous systems where autonomous devices keep communicating amongst themselves with less and less human intervention. Static or transient security mechanisms based on pure authentication fall short in this case, as every device that has nominally been put in a non-trust domain can push bad from there. OBJECTIVES: To guard against, we present in this paper a Scalable Dynamic Trust Evolution (SDTE) framework for large-scale IoT information systems. METHODS: We model trust as a dynamically changing latent state dependent on multiple sources of evidence, such as direct interaction, neighbour recommendation, context condition, and historical behaviour. A temporal update mechanism with exponential forget rate (λ) allows fast adaptation to changing behaviour, while asynchronous weighting (β⁻ > β⁺) indicates negative actions have larger weighting on trust than positive ones. An uncertainty-aware fusion of heterogeneous evidence sources mitigates conflicting observations. Overall, the framework is fully distributed and operates in lightweight computation (O(d)) without any need for sufficient computing resources, and can thus be deployed on low-resource edge devices. RESULTS: Our experimental results on publicly available IoT datasets (IoT-23 and Edge-IIoTset) show that SDTE achieves high detection performance (96.8% accuracy, 95.7% F1-score) across malicious ratio scenarios, good robustness against intermittent on–off attacks, and scaling as the network grows in size. CONCLUSION: The proposed SDTE framework provides an effective, lightweight, and scalable trust management solution for large-scale IoT information systems.