Inspiration
Online scams and phishing attacks are growing rapidly, targeting both individuals and organizations. We wanted to build a tool that helps people detect phishing websites in real time using the power of AI and automation.
What it does
PhishGuard analyzes:
🌐 URLs → checks structure, reputation, and patterns using ML models. 🖼️ Screenshots → detects fake website layouts via CNN-based image classification. 👥 User feedback → improves detection accuracy by learning from real-world reports. It provides users with a fast, reliable, and easy-to-use extension/app that flags dangerous websites before they can cause harm.
How we built it
Backend (Flask/Node.js) → hosts ML models and handles API requests. Frontend (React.js/Chrome Extension) → user interface for real-time detection. Machine Learning Models: BERT & Random Forest → URL phishing detection. CNN + OpenCV + OCR → Screenshot-based phishing detection. Database (MySQL/PostgreSQL) → storing results & user feedback. Deployment → hosted on cloud (Heroku/Render/Vercel).
Challenges we ran into
False positives/negatives during ML training. Handling real-world edge cases like double slashes in URLs (//wp-content//secure.php). Deploying ML models efficiently with low latency.
Accomplishments that we're proud of
Achieved high accuracy in detecting phishing sites. Built an end-to-end system (ML models + Chrome Extension + Web app). Designed a solution that can genuinely help people stay safe online.
What we learned
How phishing attacks evolve and bypass basic filters. Practical experience in combining ML + Image Processing + Web Development. Improved teamwork, version control, and deployment skills.
What's next for PhishGuard
Mobile app version (Android using Kotlin/Flutter). Real-time blacklist integration (Google Safe Browsing API). Gamified phishing-awareness training for users.
Built With
- c++-frameworks:-flask
- express.js-ml/dl:-scikit-learn
- javascript
- languages:-python
- postgresql-platforms-&-tools:-github
- pytorch-image-processing:-opencv
- react.js
- replit
- tensorflow
- tesseract-ocr-databases:-mysql
- vercel
- vs-code
