Deep Social Link Prediction in the Metaverse

In the Metaverse, trust relationships are essential for building genuine connections and ensuring the credibility of shared information. Traditional methods, such as graph attention networks, often treat trust interactions as fixed, neglecting their fluid nature. This reduces the effectiveness of trust modeling and hinders the understanding of how trust evolves in virtual spaces. We introduce Tformer, a new approach that incorporates temporal trust interactions to predict and monitor the development of trust within the Metaverse. By embedding dynamic topologies based on user interactions, Tformer captures evolving trust features that shift over time. It uses temporal encoding within a multi-faceted attention network to evaluate the relevance of these changing trust characteristics. Additionally, it applies reciprocal interaction transfer to model shared trust features across different virtual environments. Tformer improves our ability to model and comprehend the evolution of trust in the dynamic, immersive landscape of the Metaverse.