Multi-view Trust based Team Recruitment for Collaborative Crowdsensing

Abstract

Collaborative crowdsensing (CCS) requires the recruitment of teams to accomplish complex sensing tasks, where trust among members plays a critical role in team performance. However, current team recruitment schemes often focus on a single type of trust relationship, overlooking the diversity of trust within teams and limiting dynamic member adjustments based on task requirements. Therefore, incorporating multi-view trust relationships into collaborative team recruitment is of great importance. To address this, we propose a novel Multi-view Trust based Team Recruitment (MTrust-TR) scheme for CCS. First, we model multi-view trust networks for CCS to describe the multiple trust relationships among users. Next, for large-scale crowdsensing platforms, we propose SGAT-Trust, a scalable trust evaluation model based on graph attention networks, which efficiently evaluates multi-view trust relationships through subgraph extraction and attention coefficient assignment, significantly reducing computational overhead. Then, leveraging these trust relationships, we further develop a graph learning-based recruitment method that dynamically integrates multi-view trust relationships according to task requirements, enabling the formation of optimal collaborative teams tailored to diverse tasks. Experimental results demonstrate that the SGAT-Trust model achieves significant advantages in memory consumption and running time, and the proposed recruitment scheme effectively improves the quality of service (QoS) in CCS.