Evading User-Specific Offensive Web Pages via Large-Scale Collaborations

Web pages polluted by unhealthy contents (e.g. pornography or violence) have offended many users and become a social headache. This paper presents a collaborative rating system and a light-weight algorithm to detect polluted pages and thus improve user experience of web browsing. It mainly tackles two challenges. First, the system should cater to web users´ different tastes and judging standards on which polluted pages they like or dislike. Second, the system should be resilient to dishonest ratings and collusions. The model and the algorithm are evaluated by simulations which show that they can work well.