Abstract :
Abstract The usage of social media platforms has grown, offering individuals diverse avenues for communication, expressing opinions and sharing online content. However, this surge has also given rise to the emergence of social bots, which are programmed accounts designed to imitate human behavior. Such bots possess the capability to disseminate false information, manipulate financial markets, aid terrorism, and disrupt democratic processes. To tackle this issue, various approaches have been utilized to detect social bots, including approaches based on profiles, time patterns, content analysis, behavior, and network characteristics. However, neither of the approaches effectively combines all these features to implement social bot detection comprehensively. This paper introduces an ensemble methodology that merges profile, behavioral, temporal, network, graph, and content-based attributes, culminating in a comprehensive model for discerning social bots on the Twitter platform. We utilize the Twibot-22 dataset for conducting experiments and evaluate the performance of our approach against benchmark models. The XGBoost model, with an accuracy of 0.898, exhibited superior performance compared to the benchmark models. This research contributes to the continuous endeavor focused on safeguarding the authenticity of tweet content and mitigating the risks associated with social bots on social networks.
Keyword :
N/A