Challenges in machine learning-based social bot detection: a systematic review

The rise of social digital network platforms has led to the widespread presence of social bots, automated entities that actively engage with users and content. While some bots serve beneficial purposes, others pose significant threats by manipulating public opinion, spreading misinformation, and conducting coordinated attacks. As a result, detecting and mitigating the influence of social bots has become a critical area of research. Machine learning-based detection techniques have shown promise in distinguishing bots from human users; however, their implementation presents several challenges, including issues related to data quality, model generalizability, feature selection, adversarial resistance, and computational efficiency. This systematic literature review aims to identify, categorize, and analyze the key challenges obstructing the advancement of machine learning-based social bot detection systems while consolidating existing strategies proposed in the literature to address these obstacles. The review follows Kitchenham and Charters’ SLR methodology. Four major academic databases were searched (Springer, IEEE Xplore, ScienceDirect, and ACM Digital Library) using a structured Boolean queries. By synthesizing findings from studies published between January 2010 and June 2024, this review provides a comprehensive overview of the limitations in current detection approaches and offers insights that can guide future advancements in the field. The findings of this study aim to support researchers and practitioners in improving the robustness and effectiveness of social bot detection methodologies.