Detection of Malicious Social Bots Using Learning Automata with URL Features in Twitter Network

Abstract

With the rapid growth of social media platforms like Twitter, malicious social bots have become a significant threat, capable of manipulating public opinion, spreading misinformation, and launching cyber-attacks. These bots often mimic human behavior, making their detection a challenging task. This project proposes a novel approach to detect malicious social bots on Twitter by leveraging Learning Automata in combination with URL-based features extracted from user-generated content. The methodology involves analyzing embedded URLs in tweets—such as domain reputation, frequency, and redirection patterns—to identify abnormal behavior commonly associated with bot activity. Learning Automata, a reinforcement learning technique, is employed to adaptively improve the classification of accounts as benign or malicious over time. The proposed model is trained and evaluated on real-world Twitter datasets, demonstrating improved accuracy and adaptability compared to traditional machine learning classifiers. This work contributes to the ongoing efforts in social media security by offering a dynamic, scalable, and feature-rich solution for identifying and mitigating the threat posed by malicious social bots.