Detection of fake Twitter followers using graph centrality measures

With the exponential rise in the number of Internet users, Social Networking platforms have become one of the major means of communication all over the globe. Many major players in this field exist including the likes of Facebook, Twitter, Google+ etc. Impressed by the number of users an individual can reach using the existing Social Networking platforms, most organizations and celebrities make use of them to keep in touch with their fans and followers continuously. Social Networking platforms also allow the organizations or celebrities to publicize any event and to update information regarding their business just to keep themselves active in the market. Most Social Networking platforms provide some form of metric which can be used to define the popularity of an user such as the number of followers on Twitter, number of likes on Facebook etc. However, in recent years, it has been observed that many users attempt to manipulate their popularity metric with the help of fake accounts to look more popular. In this paper, we have devised a method which can be used to detect all the fake followers within a social graph network based on features related to the centrality of all the nodes in the graph and training a classifier based on a subset of the data. Using only graph based centrality measures, the proposed method yielded very high accuracy on fake follower detection. The proposed method is generic in nature and can be used irrespective of the social network platform under consideration.