Accurate Trust Evaluation for Effective Operation of Social IoT Systems via Hypergraph-Enabled Self-Supervised Contrastive Learning

Social Internet-of-Things (IoT) enhances collaboration between devices by endowing IoT systems with social attributes. However, calculating trust between devices based on complex and dynamic social attributes—similar to trust formation mechanisms in human society—poses a significant challenge. To address this issue, this paper presents a new hypergraph-enabled selfsupervised contrastive learning (HSCL) method to accurately determine trust values between devices. To implement the proposed HSCL, hypergraphs are first used to discover and represent high-order relationships based on social attributes. Hypergraph augmentation is then applied to enhance the semantics of the generated social hypergraph, followed by the use of a parametersharing hypergraph neural network to nonlinearly fuse the high-order social relationships. Additionally, a self-supervised contrastive learning method is utilized to obtain meaningful device embeddings by conducting comparisons among devices, hyperedges, and device-to-hyperedge relationships. Finally, trust values between devices are calculated based on device embeddings that encapsulate high-order social relationships. Extensive experiments reveal that the proposed HSCL method outperforms baseline algorithms in effectively distinguishing between trusted and untrusted nodes and identifying the most trusted node.