Multimodal Counterfeit Intelligence Framework for Authenticity Verification in Online Market Places

Syed Jasmin1, C Rajesh2, K Vishnupriya 3, K M Doraswamy 4, N Rohith Kumar5
1,2,3,4,5 Computer Science and Information Technology, Siddharth Institute of Engineering & Technology

ABSTRACT - This project addresses the erosion of consumer trust caused by counterfeit e-commerce listings by developing a multi-modal detection system that identifies inconsistencies acrosstextual descriptions, product images, and seller metadata. The technical framework utilizes transformer-basedembeddings to flag linguistic anomalies and CNN/ViT  encoders to compare listing images against authenticbrand references. By integrating these features through both early and late fusion strategies, the system generates a risk-prioritized likelihoodscore accompanied by explanatory highlights for human reviewers. Validated against adversarial examples and evaluated using metrics such as F1-score and false positive rates, the solution supports scalable deployment through batch and streaming APIs. The resulting interactive dashboard provides a robust defense mechanism for online marketplaces. Keywords: Multi-Modal Deep Learning, Fake Product Detection, Transformer Models, CNN (convolutional Neural Networks), Vision Transformers (ViT), Feature Fusion, Seller Metadata Analysis.