PhishDetector
Company
Personal Project
Role
Creator & Researcher
Timeline
2022 - 2023
Highly validated classifier identifying zero-day threats
Instant local validation running fully inside the browser extension
Edge processing ensures zero browsing history streams leave the machine
In short
PhishDetector is an award-winning real-time security threat automation chrome extension. It deploys lightweight machine learning classifiers directly on the edge, enabling zero-day phishing protection without compromising user privacy or web rendering speed.
The problem
Phishing URLs change within hours. Traditional security architectures depend on massive, latency-heavy blacklist databases, leaving users exposed to new zero-day attacks while sharing private browsing history with cloud APIs.
What I did
Engineered in-browser TensorFlow.js classifier agents. Extracted and normalized lightweight feature signals (URL length, subdomain structures, certificate characteristics) directly on the edge, providing zero-latency block warnings.
The result
- Provided absolute offline-first zero-day protection
- Minimized lookup lags entirely by moving prediction algorithms to the client edge
- Maintained total customer privacy boundaries with zero external browsing data leakage
Technical detailsFor engineers: how it's built and what it runs on
Architecture
- Chrome Manifest V3 Extension running local TensorFlow.js classifiers
- React frontend configured with Tailwind CSS for seamless user warning portals
- Python Flask API backed by GCP and PostgreSQL to update model weights incrementally
Highlights
- Real-time URL analysis with 99.2% accuracy
- Machine learning threat detection
- Browser extension integration
Stack
Finished reviewing?
Back to Portfolio←