Detecting Fake Accounts with Multi-Platform ML

Maintaining user trust and digital security is severely hampered by the growing number of fake accounts on social media sites. This paper aims to improve platform integrity by creating a reliable system that can identify fake accounts using advanced machine learning techniques. Traditional methods for detecting such accounts have relied heavily on heuristic-based strategies. These methods lack the scalability and adaptability required to cope with evolving fraudulent activities. The proposed Detecting Fake Accounts with Multi-Platform ML (DFMML) framework overcomes existing limitations by leveraging diverse datasets from Instagram, Facebook, and X (formerly Twitter) to train multiple machine learning models, including AdaBoost, Bagging, Decision Trees, Logistic Regression, Random Forest, and Gradient Boosting. A user-friendly Streamlit application has also been built, offering real-time detection results with dedicated pages for each platform.