ReSpam: A Novel Reputation Based Mechanism of Defending against Tag Spam in Social Computing

Abstract

Defending against spam in tagging system is a verychallenging task. This paper presents ReSpam, a novel reputationbased mechanism of defending against tag spam in taggingsystems. In ReSpam, timing behaviors and voting methods areintroduced to defend tag spam. Each user has a global reputationwhich can increase or decrease after some judgement given bythe tagging system. When the system finds that the time betweenlogging and posting is too short or the time between opening awebpage and posting a tag is too short, the reputations of all theusers who post a tag of that kind will decrease. For a uncertaintag, the system requests some users to help to judge the quality ofthat uncertain tag. If the feedback shows that lots of users regardthat tag as spam, all the users who post a tag of that kind willfind their reputations lower than before. Otherwise, those users' reputations will increase. The advantage of ReSpam is its simplerationale and less computing compared to pairwise reputationbased mechanism such as SpamClean and DSpam. The systemranks search result by the average reputation of annotators ofeach result. Experimental results show ReSpam can effectivelyresist tag spam and work better than some existing tag search schemes.