Towards a Reliable Spam-Proof Tagging System

Abstract

Tagging systems are particularly vulnerable to tag spam. Although some previous efforts aim to address this problem with detection-based or demotion-based approaches, tricky attacks launched by attackers who can exploit vulnerabilities of spam-resistant mechanisms are still able to invalidate those efforts. Therefore, it is challenging to resist tricky spam attacks in tagging systems. This paper proposes a novel spam-proof tagging system, which can provide high-quality tag search results even under tricky attacks, based on four key insights: demotion-based strategy, reputation, altruistic users and social networking. Specifically, our system upgrades/degrades the ranks of correct/incorrect content items in search results through introducing personalized users' reliability degrees and responsible users, thus avoiding clients pick unwanted content. Experimental results illustrated our system could effectively defend against tricky tag spam attacks and work better than current prevalent tag search models.