Detection of Malicious Social Bots using Learning Automata with URL Features in Twitter Networks

Abstract

This research investigates the detection of malicious social bots on Twitter using a novel approach that leverages Learning Automata (LA) and URL features extracted from users' tweets. The increasing prevalence of malicious bots necessitates robust detection mechanisms. This study explores the effectiveness of LA in classifying users based on their URL patterns, addressing the limitations of existing rule-based and machine learning methods. The results demonstrate improved accuracy and efficiency in identifying malicious bot accounts compared to baseline techniques, contributing to a more secure and authentic Twitter environment.

Introduction

The proliferation of malicious social bots on platforms like Twitter poses significant threats, including spreading misinformation, manipulating public opinion, and conducting coordinated attacks. Existing bot detection methods often rely on handcrafted features or struggle with evolving bot behavior. Furthermore, URL features, reflecting a user's online activity and potential malicious intent, remain under-explored in bot detection. This research addresses this gap by proposing a Learning Automata-based approach that effectively utilizes URL features for enhanced malicious bot detection on Twitter, contributing to a safer online environment.

Objectives

• To develop a robust and efficient bot detection system using Learning Automata.
• To evaluate the effectiveness of URL features in improving bot detection accuracy.
• To compare the performance of the proposed system with existing bot detection methods.