Tag quantification for spam detection in social bookmarking system

In this paper, we describes spam detection, based on the analysis of posts, in social bookmarking sites. For real-time detection of spam posts, we suggest a tag quantification scheme and a selective evaluation method for choosing tags. The tag quantification scores every tag. In the selective evaluation, the tag scores based on the usage frequency and the proportion of spammers are measured and the concepts of white tag and black tag are introduced. Using these concepts, tags are systematically categorized into the tags hindering the performance of spam detection, the tags helpful in capturing spammers, and the tags which should incur a penalty. Finally, we suggest semantic features to further improve the spam detection.