Crowdsourcing aims to aggregate collective intelligence from independent work of individuals using online platforms. However, some participants disregard platform regulations for higher payment, collaborating overtly or covertly to reduce individual workload—a behavior known as collusion in crowdsourcing. Prior research on collusion has focused on small groups or offline formations, neglecting the potential for large-scale anonymous collusion. Our study unveils a feasible method for large-scale anonymous collusion, making current detection methods less effective. In this paper, we propose a novel framework for large-scale collusion, demonstrating its negative impacts. Workers using this framework can earn more commissions with reduced workload, while at the same time easily evading current detection systems. Meanwhile, to address such collusive behavior, we present a graph-oriented collusion detection algorithm that can successfully identifies colluding workers. Experimental results show that our method outperforms previous ones, significantly reducing adverse effects on crowdsourced data by excluding collusion cases.
Read full abstract