Revealing, characterizing, and detecting crowdsourcing spammers: A case study in community Q&A
Aifang Xu, Xiaonan Feng, Ye Tian · 2015
Crowdsourcing services have emerged and become popular on the Internet in recent years. However, evidence shows that crowdsourcing can be maliciously manipulated. In this paper, we focus on the “dark side” of the crowdsourcing services. More specifically, we investigate the spam campaigns that are originated and orchestrated on a large Chinese-based crowdsourcing website, namely ZhuBaJie.com, and track the crowd workers to their spamming behaviors on Baidu Zhidao, the largest community-based question answering (QA) site in China. By linking the spam campaigns, workers, spammer accounts, and spamming behaviors together, we are able to reveal the entire ecosystem that underlies the crowdsourcing spam attacks. We present a comprehensive and insightful analysis of the ecosystem from multiple perspectives, including the scale and scope of the spam attacks, Sybil accounts and colluding strategy employed by the spammers, workers' efforts and monetary rewards, and quality control performed by the spam campaigners, etc. We also analyze the behavioral discrepancies between the spammer accounts and the legitimate users in community QA, and present methodologies for detecting the spammers based on our understandings on the crowdsourcing spam ecosystem.