Multi-Modal Fake Profile Detection Using Text, Image, and Behavioral Analysis

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.