Defending against Spam in Tagging Systems via Reputations

Abstract

"Global Internet is witnessing a rapidly growing popularity of tagging services on the social networks, which enable people to share and tag different categories of resources. However, the current tagging systems face a serious problem -- tag spam. In this paper, we propose SpamLimit -- a novel social- enhanced reputation mechanism against spam in tagging systems. First, we propose a basic reputation mechanism that provides the personalized reputation estimates to each user in system. Our approach can impose severe and quick punishment to spammers but also provide an incentive to promote normal users sharing the correct tags. Because users can rank the tag search results with the reputation estimates of owners of resources, the results provided by spammers can be degraded to the end of search results. Then, we utilize friend relationships, the social nature of tagging systems, to enhance the basic reputation mechanism. Because the friends are all real-world acquaintances, these reliable companions can provide many referential experiences to users. This will help to improve both performance and convergence of SpamLimit. Finally, our experiment results illustrate that SpamLimit can effectively defend against tag spam and work better than the existing tag search models in tagging systems."