XSS Detection Using CNN (Deep Learning)

This project demonstrates how to detect Cross-Site Scripting (XSS) attacks in HTML using a Convolutional Neural Network (CNN). It converts HTML content into ASCII matrices and feeds them to a CNN for binary classification (malicious vs. benign).

🔍 Features

• Converts HTML code into normalized ASCII matrices (image-like format)
• Classifies input as XSS or benign using a trained CNN model
• Preprocessing pipeline with BeautifulSoup and OpenCV
• Model trained on labeled dataset from Kaggle

📁 Project Structure

 - XSS-Detection/
     -     ├── model/ # Trained model
     -     ├── data/ # Dataset
     -     ├── test_html/ # Sample HTML files
     -     ├── src/ # Source code
     -     ├── README.md # This File
     -     ├── ProjectReport.pdf # Detailed report of project

---

🚀 How It Works

1. HTML Preprocessing:

   • HTML is cleaned using BeautifulSoup
   • Each character is converted to an ASCII value
   • The ASCII array is padded and reshaped to a 100x100 matrix
   • Matrix is normalized to values between 0 and 1

2. CNN Architecture:

   • 3 Convolutional layers with ReLU + MaxPooling
   • Dense layers for decision-making
   • Sigmoid output for binary classification

3. Prediction:

   • Model outputs a probability between 0 and 1
   • Threshold > 0.5 → XSS detected

---

🔍 Example Predictions

HTML Snippet	Prediction	Confidence
<a href="http://example.com">Click</a>	✅ Safe	0.04
<script>alert('XSS')</script>	❌ XSS Detected	0.98
<img src="x" onerror="alert('XSS')">	❌ XSS Detected	0.92
<div onclick="console.log('safe')">Click</div>	✅ Safe	0.12

---

🧪 How to Test

---

1. Clone the repo:

git clone https://github.com/IJBaig/XSS-Detection.git
cd xss-detection

2. Place your HTML file in the test/ directory.

3. Run the Python test script:

python src/test_xss_detection.py

---

📝 Jupyter Notebook

The training process is implemented entirely in a Jupyter Notebook:
📁 src/train_model.ipynb

🧰 Libraries Used

• TensorFlow / Keras
• OpenCV
• NumPy / Pandas
• Matplotlib

📌 Notebook Includes

• Data preprocessing
• Model architecture definition
• Training and validation visualization
• Model saving

---

📄 Report

For a detailed explanation of the project's methodology, implementation, and results, refer to:
📁 📄 Full Project Report