SpamClean: Towards Spam-Free Tagging Systems

Abstract

Tagging systems are known to be particularly vulnerable to tag spam. This paper introduces SpamClean, a novel social experience-based scheme, and presents the performance of SpamClean to defend against the tag spam in tagging systems. We first propose a novel mechanism based on cosine technique to compute the correlations between the client and other users in the system, and look the correlations as the experiences of the client with respect to other users. The client ranks each tag search result based on the average of experiences of the client with respect to all the owners of this result. To obtain higher quality search results, we propose socially-enhanced mechanism - using the friend-relationships, the social nature of tagging systems, to enhance SpamClean. This is based on considering that the client's social friends can share their previous experiences and help improve both the performance and convergence of SpamClean. Finally, the experimental results illustrate that SpamClean can effectively defend against tag spam and work better than the existing search models in the current tagging systems.