Identification of Anomalous Users in Twitter based on User Behaviour using Artificial Neural Networks

Abstract

Increased number of users in Online Social Networks(OSN) such as Facebook, Twitter, Instagram etc has led to privacy and security concerns. In this paper, we propose a behavioral based risk assessment method which will help to identity Social Bots and Compromised Accounts in Twitter by leveraging the Neural Networks. Our solution can detect unusual accounts based on the users' writing style, such as, time of tweeting, language used for writing tweet, frequency of tweeting, etc. This way, if the writing style of the user changes from what is considered to be normal, we flag the user as risky and the OSN admin can formulate strategies to overcome the same. Experiments were performed on the real world dataset extracted from twitter. Results show tremendous accuracy in identifying Social Bots and that only few tweets are required to identify whether an account has been compromised or not.