Toward Efficient Spammers Gathering in Twitter Social Networks

Abstract

This paper introduces a novel system, named pseudo-honeypot, for efficient spammers gathering. Different from the manual setup in the honeypot, the pseudo-honeypot takes advantage of Twitter users' diversity and selects accounts with the attributes of having the higher potentials of attracting spammers, as the parasitic bodies. By harnessing a set of normal accounts possessing these attributes and monitoring their streaming posts and behavioral patterns, the pseudo-honeypot can gather the tweets that are far more likely of including spammer activities, while removing the risks of being recognized by smart spammers. It substantially advances the honeypot-based solutions in attribute availability, deployment flexibility, network scalability, and system portability. We present the system design and implementation of pseudo-honeypot (including node selection, monitoring, feature extraction, and learning-based classification) in Twitter networks. Through experiments, we demonstrate its effectiveness in term of spammer gathering.