Spammer detection based on Hidden Markov Model in micro-blogging

Over the past decade, social network becomes the main way for people to obtain information and communicate with their online friends. With the tremendous increase of social network, spammer also develops rapidly and spam information posted by spammer seriously affects people's daily life. Although the official executes some strict strategies for prevention, spammer evolves dynamically to adapt these strategies at the same time. They disguise as common users with posting interesting tweets and posting spam with irregular time. To deal with this problem, we present a method for spammer detection on the basis of Hidden Markov Model in micro-blogging. The proposed method classifies the user into spammer and non-spammer according to their major count-based features and behavior-based features. The usage of HMM allows us to describe the user behavior with hidden states dynamically, which can be decoded from several observed features. The experimental results based on real data collected from Sina micro-blogging site demonstrate the effectiveness of our method.