Phishing Detection Web Application (ML-Powered) — Project
Built an ML-powered phishing detection web application that labels URLs as malicious or benign in real time. He curated training datasets and engineered features, then evaluated model outputs against ground-truth labels to measure accuracy and coverage. The work required identifying edge cases and validating performance using an external safety reference for correctness checks. • URL malicious/benign classification labeling workflow • Training dataset curation and feature engineering decisions • Output evaluation vs. ground-truth labels • Edge-case assessment using Safe Browsing API