A Deep bidirectional transformer based Twitter spam detection and profiling

Abstract

Online social networks are becoming extremely popular among Internet users as they spend a significant amount of time on popular social networking sites like Facebook, Twitter, and Google+. These sites are turning out to be fundamentally pervasive and are developing a communication channel for billions of users. Twitter has become a target platform on which spammers spread large amounts of harmful information. These malicious spamming activities have seriously threatened normal users’ personal privacy and information security. An effective method for detecting spammers is to learn about user features and social network information. However, social spammers often change their spamming strategies for evading the detection system. To tackle this challenge, in this research we determine various features to capture the consistency of users’ behavior. In this research, we investigate additional criteria – spam patterns, to measure the similarity across accounts on Twitter. We propose a method to define the relation among accounts by investigating their tweeting patterns and content. Our real data evaluation reveals that, given some initially labelled spam tweets, this approach can detect additional spam tweets and spam accounts that are correlated to the initially labelled spam tweets.