Sybil Defense in Crowdsourcing Platforms

摘要: Crowdsourcing platforms have been widely deployed to solve many computer-hard problems, e.g., image recognition and entity resolution. Quality control is an important issue in crowdsourcing, which has extensively addressed by existing quality-control algorithms, voting-based algorithms probabilistic graphical models. However, these cannot ensure quality under sybil attacks, leverages a large number of accounts generate results for dominating answers normal workers. To address this problem, we propose defense framework can help crowdsourcing identify workers the attack. We develop similarity function quantify worker similarity. Based on similarity, cluster into different groups such that utilize small golden questions accurately groups. also devise online instantly detect throttle attacks. Our method ability multi-attackers one task. best our knowledge, first crowdsourcing. Experimental real-world datasets demonstrate effectively