Scholarlink

<p>Social trust is a crucial concept in online social networks (OSNs), serving as the foundation for social networks applications ranging from making friends, e-commerce, and network security. To date, most existing approaches focus on introducing functional time encoding or resorting to the evolution of node embedding information to extract the temporal feature of social trust in dynamic OSNs. However, this method fails to accurately assess the trustworthiness between two users in time-varying OSNs with varying user sets. In this paper, we present a novel EvolveGCN-based solution <i>EvolveTrust</i> to evaluate trust relationships in time-varying OSNs. The approach adopts graph neural networks DTNN to capture the latent factors related to social trust. Furthermore, the solution uses a gated recurrent unit(GRU) model to update the weight matrix of DTNN, which can obtain the dynamic feature of social trust instead of relying on node embedding evolution. Thus, the solution can accurately infer the trust level between users in time-aware OSNs despite the user set frequently varying. Experiment results show that our proposed solution can enhance the precision of trust assessment in time-aware OSNs compared with benchmark counterparts, even if user sets frequently vary.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

EvolveTrust: evolving graph neural network for trust assessment in time-aware social networks

  • 期刊论文

  • 发表日期：

    2026年2月17日

查看全文
  • Jie Wen,
  • Nan Jiang,
  • Hualin Zhan

摘要

Social trust is a crucial concept in online social networks (OSNs), serving as the foundation for social networks applications ranging from making friends, e-commerce, and network security. To date, most existing approaches focus on introducing functional time encoding or resorting to the evolution of node embedding information to extract the temporal feature of social trust in dynamic OSNs. However, this method fails to accurately assess the trustworthiness between two users in time-varying OSNs with varying user sets. In this paper, we present a novel EvolveGCN-based solution EvolveTrust to evaluate trust relationships in time-varying OSNs. The approach adopts graph neural networks DTNN to capture the latent factors related to social trust. Furthermore, the solution uses a gated recurrent unit(GRU) model to update the weight matrix of DTNN, which can obtain the dynamic feature of social trust instead of relying on node embedding evolution. Thus, the solution can accurately infer the trust level between users in time-aware OSNs despite the user set frequently varying. Experiment results show that our proposed solution can enhance the precision of trust assessment in time-aware OSNs compared with benchmark counterparts, even if user sets frequently vary.