QGuard is a real-time cybersecurity tool designed to detect and block "Quishing" (QR Code Phishing) attacks. It uses a Hybrid Intelligence System combining Machine Learning (Random Forest) with rule-based whitelisting to verify QR codes before the user opens them.
QR codes are everywhere—payments, menus, and logins. However, humans cannot read QR codes, making them a perfect vehicle for cyberattacks. Hackers embed malicious URLs inside QR codes, bypassing traditional email filters.
QGuard acts as a "security guard" between the camera and the browser. It scans the QR code, analyzes the URL structure using AI, and provides an instant safety rating:
- 🟢 Green: Safe Website.
- 🟠 Orange: Verified Payment (UPI, GPay, Paytm).
- 🔴 Red: Danger / Malicious Phishing Pattern.
- 📸 Real-Time Scanning: Live video feed detection using OpenCV.
- 🧠 AI Analysis: A Random Forest Classifier trained on 15,000+ malicious and benign URLs.
- ✅ Payment Whitelist: Frictionless verification for trusted payment apps (
upi://,gpay,paytm). - 🛡️ Hybrid Detection: Combines AI probability with rigid "Blacklist Rules" to catch specific threats like
verify-accountorupdate-password. - ✍️ Manual Link Checker: A dedicated tab to manually paste and analyze suspicious SMS/Email links.
- 🔒 Privacy First: Runs 100% offline. No data leaves your device.
- Language: Python
- Interface: Streamlit (Web UI)
- Computer Vision: OpenCV (
cv2) - Machine Learning: Scikit-Learn (
sklearn) - Data Handling: Pandas & Numpy
The AI model (phishing_model.pkl) is included in this repository, so the application runs out-of-the-box.
However, the training dataset (data.csv) is not included due to GitHub's file size limits. If you wish to retrain the model yourself:
- Download the dataset from Kaggle: Phishing Site URLs Dataset.
- Rename the downloaded file to
data.csv. - Place it in the project folder.
- Run
python train_model.py.
git clone [https://github.com/CodeWhizAnu/QGuard.git] (https://github.com/CodeWhizAnu/QGuard.git))
cd QuishGuard-QR-Security
You can install all required libraries with this single command:
pip install streamlit opencv-python pandas scikit-learn numpy
Start the Streamlit server:
streamlit run app.py
The application will open automatically in your default web browser.
-
Input: The system captures a QR code via webcam or accepts a manual URL input.
-
Step 1 (Whitelist Check): It checks if the link is a known payment provider (e.g.,
upi://). If yes Safe (Orange). -
Step 2 (Blacklist Trap): It checks for obvious phishing keywords (e.g.,
secure-login,verify-bank). If found Danger (Red). -
Step 3 (AI Prediction): If the link is unknown, the AI model analyzes 6 structural features:
- URL Length
- Count of Dots (.)
- Count of Hyphens (-)
- Presence of '@' symbol
- HTTPS check
- Numeric character ratio
- Output: The user receives a clear visual alert (Green/Orange/Red).
- Mobile App: Convert the logic into a native Android/iOS app using Flutter or Kivy.
- API Integration: Connect with Google Safe Browsing API for double-verification.
- Browser Extension: A plugin to auto-scan QR images on websites.
Built by [Anushka Deshmukh].
