Detecting Spam Bots in Online Social Networking Sites: A Machine Learning Approach

As online social networking sites become more and more popular, the have also attracted the attentions of the spammers. In this paper, Twitter, a popular micro-blogging service, is studied as an example of spam hots detection in online social networking sites. A machine learning approach is proposed to distinguish the spam hots from normal ones. To facilitate the spam hots detection, three graph-based features, such as the number of friends and the number of followers, are extracted to explore the unique follower and friend relationships among users on Twitter. Three content-based features are also extracted from user's most recent 20 tweets. A real data set is collected from Twitter's public available information using two different methods. Evaluation experiments show that the detection system is efficient and accurate to identify spam hots in Twitter.