Social-media advertisements make product discovery easy, but the same channel can direct users to fake storefronts that steal credentials and payments. This Projects presents FraudGuard, a machine-learning system that examines a shopping landing-page URL and the measurable properties of the page behind it before the user commits to a purchase. The work uses the PhiUSIIL Phishing URL Dataset of 235,795 instances (134,850 legitimate, 100,945 phishing). Two feature sets were prepared: a 50-feature numeric benchmark and a 42-feature deployment vector that can be regenerated from a live URL. Logistic Regression, Random Forest and XGBoost were compared. On the 50-feature benchmark, Logistic Regression reached 99.98% accuracy and both tree ensembles reached 100.00% on the held-out split; XGBoost was then chosen for deployment on the 42-feature representation. A Streamlit prototype with Home, How It Works, Check Website and About pages was built around the model. Because perfect benchmark scores rarely transfer to unseen campaigns, the paper also analyses feature dependency, limitations and the external validation still required.

Prachi Khansuli, Dr. Manoj Singh, "Social Media Shopping Fraud Detection Using Machine Learning" International Journal of Scientific Research and Engineering Development, V9(5): Page(1371-1375) September - October 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.