SOCIAL SPAMMER DETECTION VIA CONVEX NONNEGATIVE MATRIX FACTORIZATION

Abstract

Given that the popularity of social network websites, such as Twitter and Sina Weibo, many criminal individuals have been referred to as social spammers, spreading unlawful information to ordinary users. Multiple methodologies are designed to identify spammers using a trained classifier via optimization techniques, especially the use of content and social following data. The development of spammers, along with the good will of some legitimate users, makes social monitoring of information prone to manipulating spammers. At the same time, potential social activities and behavior show considerable variability among users, leading to a large but thin space for current model methodologies. This research proposes a new CNMFSD approach to identify spammer in social networks and in an innovative way uses both content and user - interaction relations. Empirically, we evaluated the proposed method using the Twitter data file in the real world, and experimental results indicate that the CNMFSD method significantly increases detection performance compared to basic models