🌐 SafeSurfer: AI-Powered Web Safety Chrome Extension

⚠️ Under Active Development | 🛡️ Real-time Website Scanning | 🧠 Hybrid AI Detection | 📊 MongoDB Logging

---

🚀 Overview

SafeSurfer is a smart browser security companion that analyzes every website you visit in real-time, determines if it's safe or malicious, and alerts you instantly — all using a combination of AI models, domain intelligence, and user feedback.

---

🧠 Features

• ✅ Real-Time URL Scanning as you browse
• ⚡ XGBoost + Deep Learning Hybrid Model for phishing/malicious URL detection
• 🔐 Post-check Rules for trusted or suspicious domains
• 🧩 Chrome Extension Integration
• 📦 Flask API Backend for ML inference
• ☁️ MongoDB Logging of visited URLs & verdicts
• 📈 Future-ready for feedback-based retraining and dashboard integration

---

🧩 System Architecture

[Chrome Extension] → [Flask API] → [ML Models] → [MongoDB Storage]
         ↑               ↓
     Real-time alerts   Logs & Feedback

---

⚙️ Current Stack

Layer	Tech
Frontend	HTML/CSS/JS (Chrome extension)
Backend API	Flask (Python)
ML Models	XGBoost + TensorFlow
Storage	MongoDB (Localhost)
UI	Extension Popup + Badge Alerts

---

🔄 Flow

1. User visits a site
2. Chrome Extension captures URL
3. Sends URL to local Flask API (http://127.0.0.1:5001/predict)
4. Flask API:
   • Preprocesses URL (cleans domain, removes query)
   • Runs XGBoost + Deep Learning predictions
   • Applies rule-based overrides (e.g. whitelist)
   • Saves verdict to MongoDB
5. Verdict is shown in popup

---

🛠️ To Run Locally

1. Clone this repo
2. Start MongoDB
3. Train or load models (train_url_model.py, phase_2-model.py)
4. Run the API server:

   python url_scanner_api.py

5. Load the extension from /SafeSurfer_Chrome_Extension in Chrome (Developer Mode)
6. Visit websites and check the extension popup verdict

---

🧪 Dev Notes

• This is a development-phase project and accuracy may vary
• Model is currently trained on basic phishing and top-1M domains — needs fine-tuning for complex real-world traffic
• Feedback logging is enabled for future retraining

---

📌 Coming Soon

• 🌍 Web dashboard for browsing history analytics
• ✍️ User feedback on verdicts (used for active retraining)
• 📈 Improved model with feature engineering and full URL handling
• 🔒 OAuth-secured access and user-based dashboards

---

🤝 Contributors

Made with 💙 by @abizer007

---