Social Bot Detection Using "Features Fusion"

Abstract

Social robots are fake users who are active in social networks. Usually, social robots imitate real users and publish a lot of tweets. Malicious social robots are destroying the security of social networks. Existing research uses machine learning technology to detect social robots from social users' attributes and tweets. The features extracted from tweets include emotional features, thematic features, etc. However, the redesigned intelligent robot can imitate a certain feature of normal users' tweets, which makes the detection of social robots more difficult. The purpose of this paper is to propose several methods of fusing the old features to obtain more complex features, which makes the design of social robots more difficult. The results show that the new feature fusion method has better accuracy than the single feature for social robot detection.