Detection of Social Botnet using a Trust Model based on Spam Content in Twitter Network

Abstract

A social botnet is a collection of social bots in an online social network. The social bots are capable of performing malicious activities, such as spreading malware, phishing online social websites and posting spam content. Moreover, a social bot may generate fake messages by manipulating the belief of a legitimate participant. Hence, this leads to the online social network suffering from several vulnerabilities. In this paper, we propose a social botnet detection algorithm by incorporating a trust model (which consists of two parameters, such as direct trust and indirect trust) for identifying a trustworthy path in the online social network (like Twitter). Further, the trust value from direct relationship among participants (i.e., direct trust) is determined using Bayesian theory and the trust value from neighboring participants (i.e., indirect trust) is determined using Dempster-Shafer theory. By integrating these two parameters, trust accuracy has been improved for detecting social bots among participants. Experimentation has been done using The Fake Project dataset (collected from Twitter) to demonstrate the efficacy of the proposed social botnet detection algorithm.