Sybil-resistant Truth Discovery in Crowdsourcing by Exploiting the Long-tail Effect

Crowdsourcing refers to the employment of workers to complete tasks. Workers will also be rewarded after completing the tasks, which creates two problems. One is how to effectively aggregate the right answers of tasks when the accuracy of workers’ answers varies. The other is how to deal with Sybil attacker, who will get rewards by controlling half or more Sybil workers to influence the Truth Inference. We propose a Truth Discovery with intelligent Sybil defense, called STDEL (Sybil-resistant Truth Discovery by Exploiting the Long-tail Effect). This algorithm can improve the accuracy of Truth aggregation and reduce the rewards by identifying and banning intelligent Sybil workers. Our Experiments on two real world datasets show that STDEL has better performance on accuracy and rewards compared with state-of-the-art Sybil defense Truth Discovery.