earticle

영어

Spamming on microblogging platforms has been a critical issue in microblog-based applications, because spamming has a significant impact on information quality and credibility. In this paper, we characterize two types of spammers on microblogging platforms, namely advertised spammers and following spammers, and then present preliminary approaches to detect these spammers. We first use a real data set to characterize the features of AS and FS in terms of various aspects such as profile, behavior, and social relationship. Specially, we introduce a new feature named duplication for FS detection, which describes the duplicated behavior of users in sharing information on microblogging platforms. We present a content-sharing graph to model the relationship between users and microblogging contents, and propose an effective algorithm to calculate the duplication feature. We run several classification methods on the characterized features to test the effectiveness of the features in AS and FS detection. The results w.r.t. precision, recall, F-measure, and ROC suggest the effectiveness of our proposed features. In particular, the duplication feature is able to improve the effectiveness of FS detection.