Phishing Website Detector is a live URL scanner. Paste a link and it returns two verdicts side by side: one from VirusTotal, which pools 70+ antivirus engines and site scanners, and one from a logistic-regression model trained on URL and domain features: HTTPS, subdomains, DNS records, redirects, iframes, form handlers. Each verdict comes with a safety score, and every scan lands in a history you can export as CSV.

I built it alone in 2024, in high school, and have kept it running since: the Next.js front end, the VirusTotal integration, the scan history, and the model, trained with scikit-learn on the PhiUSIIL dataset with SMOTE balancing (ROC AUC 0.96 on held-out data). In 2026 I folded the Flask model server into the app as a TypeScript port that runs the same regression in-process on Vercel, behind an SSRF guard for the pages it fetches. Open source under MIT.