Detection of Clickbait Content Spreaders on Online Social Networks

Abstract

Users on Online Social Networks play a pivotal role in the spread of misinformation and malicious content across these platforms. Clickbait headlines are one such malicious content that causes nuisance online. Resonating the idea of ‘Precaution is Better than Cure’, this paper focused on developing methods for the early detection of malicious clickbait content spreaders on Online Social Networks by finding User's Sharing Potential for each such malicious content user. In a billion node network, as the speed of content propagation is high, by the time they are detected to be fake or harmful, it's too late to take any recovery measures. The User's Sharing Potential of a user will help identify the potential sources/spreaders of clickbait content based on their past tendencies of sharing or publishing such information on an Online Social Network. The User's Sharing Potential metric also incorporated the effect of the influence of a user's neighborhood in the network, thus combining both user and neighborhood characteristics in determining the sharing pattern of a user. Different machine learning and deep learning models were trained for detecting clickbait posts of a user with almost ninety percent accuracy for some models. The trained classifiers and graph features were used to find the sharing potential of each user. Experiments were performed on real world datasets and the results show the efficacy of the proposed approach.