Multi-Modal Fake Profile Detection Using Text, Image, and Behavioral Analysis
International Research Journal of Modernization in Engineering Technology and Science · 2025
With the exponential rise of fake accounts on social media platforms, malicious entities are now exploiting sophisticated techniques to mimic real user behavior and deceive systems.This paper proposes a robust Multi-Modal Fake Profile Detection System that combines three powerful analytical dimensions: natural language processing (NLP), image-based deep learning, and behavioral pattern analysis.Each module processes a different data modality -such as textual content, profile images, and activity logs -and outputs an independent score.These are then fused via a weighted decision model to provide the final classification: REAL or FAKE.The model achieved 95% accuracy through ensemble fusion, outperforming individual modality performance.The system also integrates explainability using SHAP values and a Flask-based web interface for real-time prediction and visualization.The framework offers a scalable and interpretable solution to combat fake profiles across platforms.