Analysis And Indentification Of Spammers In Sina Weibo

With the popularization of Internet, micro-blogging websites are nowregarded as hot platforms for communication and information exchanging.Along with the development, various kinds of spamming behaviors ariseand threaten the community security and content quality of the micro-blogcommunities. How to identify these spammers has become an importantissue for social network researchers. Most of the now existing researchprojects are for platforms like Twitter whose conclusions and resultscannot be directly applied in Chinese micro-blogging communities likeSina Weibo. So the study of spammers in Chinese micro-bloggingcommunities is very necessary.We took Sina Weibo as the study platform and propose a new way ofdetecting spammers by detecting their spamming behaviors. We collectedspammer samples using honeypot and crawler. By setting up proactivehoneypots, we attracted114spammers. By running Weibo crawler, wecollected1,459spammer samples. We also bought “Weibo promotionservice” on Taobao.com and got8,000spammer samples. We divided thesespammers into three categories according to their behaviors: aggressiveadvertising, repeated duplicate reposting and aggressive following. Wemade a deep analysis on these spammers, compared them with legitimateusers and found out their characteristics and behavior features that can tellspammers from legitimate users. Then we trained different spammingbehavior classifiers using machine learning techniques and built thespammer identification system. We finally tested the system on newlycollected test dataset. The system detected82.06%spammers and only5.92%legitimate users were mislabeled.