SpamDia: Spammer Diagnosis in Sina Weibo Microblog

Microblogs open opportunities for social spammer accounts. These widespread spammers, are threatening for microblog services and normal users. Therefore, detecting spammers should be conducted to fight and stop them. In this paper, we propose an approach to diagnose user accounts in China’s most popular microblog site Sina Weibo. Unlike existing approaches, which can hardly discover sophisticated spammers and only give a simple conclusion as spammer or not lacking of detail information, but our work provides a more responsible way that reveals the clues to verify a spammer account by using classifier-level fusion and feature-level comparison. Distinct discriminative features are used to train basic classifiers. Then a fusion model is learned to combine the outputs of the basic classifiers and make the final prediction. Comparing basic classifiers outputs with the final prediction offers the insights of spammer identification. Experiments show that our approach significantly improves the classification performance and this approach can point out spammers’ specific spam action in a detail way that helps us strike spammers accurately.