Characterizing and automatically detecting crowdturfing in Fiverr and Twitter

摘要: As human computation on crowdsourcing systems has become popular and powerful for performing tasks, malicious users have started misusing these by posting propagating manipulated contents, targeting web services such as online social networks search engines. Recently, moved to Fiverr, a fast growing micro-task marketplace, where workers can post crowdturfing tasks (i.e., astroturfing campaigns run crowd workers) customers purchase those only $5. In this manuscript, we present comprehensive analysis of in Fiverr Twitter develop predictive models detect prevent Twitter. First, identify the most types found conduct case studies tasks. Second, build task detection classifiers filter them from becoming active marketplace. Our experimental results show that proposed classification approach effectively detects achieving 97.35 % accuracy. Third, analyze real-world impact purchasing quantifying their target site (Twitter). part analysis, current security inadequately crowdsourced manipulation, which confirms necessity our approach. Finally, characteristics paid workers, find distinguishing features between legitimate accounts, use workers. are able effectively, 99.29 %