Leveraging Machine Learning for Fraudulent Social Media Profile Detection

The rapid growth of social media platforms has led to an increase in the number of fraudulent accounts, including bots, spammers, and malicious actors that undermine user trust and platform integrity. Traditional rule-based systems are often insufficient to detect such accounts due to the evolving and deceptive nature of their behavior. This paper presents a machine learning-based approach to automatically identify fraudulent social media profiles by analyzing a combination of profile metadata, behavioral patterns, content features, and network relationships. A comprehensive dataset comprising both genuine and fraudulent user accounts was compiled and preprocessed to extract relevant features. Various machine learning models, including ensemble methods and graph-based neural networks, were trained and evaluated. Among these, the XGBoost classifier achieved the highest performance with an accuracy of 94.2% and an F1-score of 90.6%, while graph neural networks demonstrated strong capability in leveraging relational data. The proposed system also incorporates a modular architecture with a feedback loop for continuous learning, making it scalable and adaptable to emerging fraud patterns. The results highlight the effectiveness of leveraging machine learning in enhancing the detection and mitigation of fraudulent profiles, contributing to safer and more trustworthy social media ecosystems.