Website Security Analyzer 3.0

Welcome to the Advanced Website Security Analyzer! This is a friendly, Python-based tool designed to help you quickly assess the security of websites. By combining tried-and-true heuristic checks with a smart machine learning model, this tool helps you spot potentially malicious websites before they can cause trouble.

What It Does

• Heuristic Security Checks

  • HTTPS Check: Quickly tells you if the website is using secure HTTPS.
  • Suspicious Patterns: Scans the URL for red flags like common malicious patterns and keywords.
  • Domain Age Analysis: Looks up the domain's age using WHOIS data to see if it's a new, and potentially risky, site.
  • SSL Certificate Validation: Checks if the site's SSL certificate is valid.
  • HTML Content Analysis: Scans the website's code for any suspicious HTML or JavaScript.
  • OWASP Vulnerability Checks: Simulates tests for common web vulnerabilities like SQL injection or XSS.

• Machine Learning Magic

  • Uses a TensorFlow model that has been trained on real-world examples to predict if a URL might be malicious.
  • Combines the ML prediction with heuristic scores to give you an overall risk rating.
  • Offers insights on which factors influenced the ML prediction the most.

• Performance & Visual Insights

  • Runs several checks at once for faster results.
  • Creates a neat bar chart (heuristic_scores.png) to show you the scores for each security check.
  • Features an interactive CLI built with the Rich library, making the tool both fun and easy to use.

• Continuous Learning

  • You can provide feedback on the analysis results, which helps retrain and improve the machine learning model over time.

Key Improvements

1. Synthetic Dataset Generation

• Function Added: _generate_synthetic_dataset(n_samples=10000)
• Purpose: Generate realistic training data with 10,000 samples instead of the original 20 samples
• Features Generated:
  • uses_https: 1 if HTTPS, 0 if HTTP (realistic distributions for malicious/benign sites)
  • suspicious_patterns_count: Count of suspicious URL patterns (higher for malicious sites)
  • domain_age_days: Age of domain registration (newer for malicious sites)
  • uses_suspicious_tld: 1 if using suspicious TLD, 0 otherwise
  • domain_length: Length of domain name (longer for malicious sites)
  • uses_ip: 1 if URL uses IP address instead of domain, 0 otherwise
  • redirects: Number of redirects in URL chain
  • subdomains_count: Number of subdomains
  • url_length: Total length of URL

2. Enhanced Neural Network Architecture

• Previous Architecture: Simple 3-layer network (16→8→1 neurons)
• New Architecture: 6-layer network with dropout for regularization
  • Input layer (9 features)
  • Dense layer (64 neurons, ReLU activation)
  • Dropout layer (30% dropout rate)
  • Dense layer (32 neurons, ReLU activation)
  • Dropout layer (30% dropout rate)
  • Dense layer (16 neurons, ReLU activation)
  • Output layer (1 neuron, sigmoid activation)
• Benefits: More complex representations, reduced overfitting

3. Feature Scaling Implementation

• Added: StandardScaler for feature normalization
• Training: Features are standardized during model training
• Prediction: Same scaling applied during inference
• Benefit: Improved model convergence and performance

4. Improved Training Process

• Validation Split: 80% train, 20% validation
• Early Stopping: Monitors validation loss with patience=10
• Multiple Metrics: Accuracy, precision, and recall tracked
• Optimizer: Adam with learning rate 0.001
• Loss Function: Binary crossentropy

5. Enhanced Retraining Process

• Synthetic Data Combination: New training data combined with 5000 synthetic samples
• Prevents Catastrophic Forgetting: Maintains knowledge of previous patterns
• Improved Architecture: Uses same enhanced network as initial training
• Proper Scaling: Combined data is scaled with updated scaler

6. Web Frontend Interface

• Flask Application: Web-based interface for easy URL analysis
• Interactive Dashboard: Real-time results display with risk visualization
• User Feedback System: Interface to provide feedback for model improvement
• Responsive Design: Works on desktop and mobile devices

Getting Started

Installation

1. Clone the Repository:

   git clone https://github.com/Harshithpilli/WebsiteSecurityAnalzyer-3.0.git
   cd WebsiteSecurityAnalzyer-3.0

2. Install the Required Libraries:

   pip install tensorflow requests python-whois tldextract numpy pandas matplotlib beautifulsoup4 rich flask

You'll need packages like TensorFlow, Requests, WHOIS, tldextract, NumPy, Pandas, Matplotlib, BeautifulSoup4, Rich, and Flask.

3. Optional Configuration: The tool disables oneDNN optimizations for TensorFlow by setting:

os.environ["TF_ENABLE_ONEDNN_OPTS"] = "0"

How To Run

Option 1: Command Line Interface

Interactive Mode Just run the tool without any arguments, and you'll enter an easy-to-use interactive menu:

python security_analzyer_3.py

You can choose to:

• Analyze a URL
• Provide feedback to help improve the model
• Exit the tool

Option 2: Web Interface

Start the Web Application Run the following command to start the Flask web server:

python app.py

This will launch the application and open your browser to http://localhost:5000

Web Interface Features:

• Clean, responsive user interface
• Real-time URL analysis
• Visual risk score display
• Detailed heuristic checks table
• Security details listing
• User feedback options to improve the model

What the Output Looks Like

Risk Score: A final score out of 10 that tells you how risky the website might be.

Risk Level: Categorized as Low, Medium, or High Risk.

Details: A clear breakdown of what was checked and why.

Visualization: A bar chart (heuristic_scores.png) that visualizes the scores from each security check.

Inside the Code

Main Class: The WebsiteSecurityAnalyzer class is the heart of the tool. It handles all the security checks and ML predictions.

Security Checks: Each security check (like HTTPS verification or SSL certificate validation) is done by its own function, making the code modular and easy to understand.

Machine Learning: The tool trains a TensorFlow model to help predict if a URL is malicious. It can also be retrained with your feedback to get even better over time.

Web Interface: The Flask application provides a user-friendly web interface for the security analyzer, making it accessible to users without command-line experience.