Spammer Classification using Ensemble Methods over Structural Social Network Features

The overwhelming growth and popularity of online social networks is also facing the issues of spamming, which mainly leads to uncontrolled dissemination of malware/viruses, promotional ads, phishing, and scams. It also consumes large amounts of network bandwidth leading to less revenue and significant financial losses to organizations. In literature, various machine learning techniques have been extensively used to detect spam and spammers in online social networks. Most commonly, individual classifiers are learnt over content-based features extracted from users' interactions and profiles to label them as spam/spammers or legitimate. Recently, new network structure-based features have also been proposed for spammer detection task, but their significance using ensemble learning methods has not been extensively evaluated yet. In this paper, we evaluate the performance of some ensemble learning methods using community-based structural features extracted from an interaction network for the task of spammer detection in online social networks.