Detecting Malicious Tweet Bots using Machine Learning Algorithms

Abstract

Social bots are social accounts operated by computer programs in online social networks, which can execute corresponding operations based on a set of procedures. With the large increase in the number, speed and variety of user data in online social networks, attempts, have been made to collect and analyze these big data. Social bots, for example, have been used to conduct automated analytical services and provide improved quality of service to consumer. Nonetheless, malicious social bots were often used to disseminate false information, and this can have real world implications. It is important to identify, delete malicious social bots in online networks. This introduces a novel approach for detecting malicious social bots for features’ selection based on the transfer likelihood of click stream sequences and semi-supervised clustering. This approach not only analyses the likelihood of change from user interaction click sources but also considers the interactions time features. Finding from our studies on real online social network platform show that the accuracy if detection of different types of malicious social bots based on transformation likelihood of user activity click streams improves by an average of 12.8 percent compared to the method of detection based on quantitative user behavior analysis.