Abstract:
Facebook is one of the popular social networking sites among people. It was founded in Feb 2004 by Harvard sophomore Mark Zuckerberg. Initially it was only used by Harvard students, but then it was expanded and now everyone from all around the world can access. It provides many features to its user like updating status, sharing posts, videos, pictures etc. Sharing videos and photos is one of the major features of Facebook. There is a risk attached with this feature e.g. if someone access the user’s account illegally and share the post (photo/video) without users’ knowledge. In Pakistan, we ae facing same risk with some additional factors, i-e., most of the people are unaware of these risks or they have no proper mechanism to protect their account from such illegal use. Development of application for social networking sites have been less focused by many researchers. This research work focuses on detecting unauthorized usage of Facebook account using behavioural pattern. The proposed application extracts the user’s pattern from his/her past sharing behaviour and then makes a user’s model based on that pattern. All the incoming posts will be matched with that model. If some inconsistent behaviour is found, it immediately detect it and generates alert to the user. In this model, two machine learning algorithms were implemented i-e., J48 and Random Tree. To test and evaluate the proposed model, we have taken ten real users accounts and have evaluated both machine learning algorithms. Result showed that the accuracy of J48 is 97% and that of random tree is 94%.
