Poster: Who spams whom? Detecting Opinion Spammer Groups and Their Spam Targets

We investigate on detection of opinion spammer groups and their spam targets in review systems. People tend to trust reviews from top-ranked reviewers much more than those from low-ranked ones [1]. Spammers may thus artificially manipulate the ranking system to make their own reviews attract more attention from others; or to make competitors’ reviews attract less attention. In this work, we thus address two aspects of potential for collaborative opinion spamming behavior (i.e., boosting or demoting). In our approach, we focus on outgoing relationships of spammers to detect users who are artificially promoted or demoted. Previously, we revealed strong positive spam communities based upon their interaction patterns and the sentiments of those interactions [2]. In this work we further explore positive/negative spam targets of such strong communities with the sentiment analysis on outgoing relationships from the strong communities. Through extensive experiments on Amazon dataset, we show that spammers tend to have interest more in their own promotion rather than in others’ demotion.