AI-PhishGuard 🛡️

AI-PhishGuard is an AI-powered phishing email detection tool with a beautiful Streamlit dashboard. It uses Natural Language Processing (NLP) and Machine Learning to classify emails as Phishing or Safe, and provides URL extraction, batch analysis, training capability, and analytics.

🚀 Features

• Single Email Analysis: Check if an email is phishing or safe.
• Batch Analysis: Upload a CSV of multiple emails for bulk detection.
• Train Your Model: Upload a labeled dataset to retrain the AI.
• Theme Support: Dark, Light, Cyberpunk, and Matrix modes.
• Analytics Dashboard: Pie charts, bar charts, and downloadable history.

📂 Folder Structure

AI-PhishGuard/
│
├── phishing_detector.py  # Main dashboard script
├── phishing_model.pkl   # Pre-trained ML model
├── requirements.txt   # Required Python packages
├── logo.png      # App logo
├── README.md   # Documentation
└── sample_dataset.csv   # Example dataset for training & testing

🛠 Installation

1. Clone the repository:

     git clone https://github.com/yourusername/AI-PhishGuard.git
     cd AI-PhishGuard
   

2. Install Dependenices:

     pip install -r requirements.txt
   

3. Run the app:

     streamlit run phishing_detector.py
   

🚀 Usage

1. Single Email Analysis

   • Select Check Email in the sidebar.
   • Paste the email text in the input box.
   • Click Analyze Email to get results.

2. Batch Email Analysis

   • Select Batch Analysis in the sidebar.
   • Upload a CSV file containing an email_body column.
   • View predictions and download results.

3. Train a Model

   • Select Train Model in the sidebar.
   • Upload a CSV file with columns email_body and label (1 = Phishing, 0 = Safe).
   • The model will be retrained and saved as phishing_model.pkl.

4. View Analytics

   • Select History & Analytics to see:
     • Phishing vs Safe email distribution
     • URL count distribution

🗂 Dataset Format

Example sample_dataset.csv:

email_body	label
Dear user, your account will be suspended unless you verify now at http://phishy-bank-login.com	1
Meeting scheduled for tomorrow at 10 AM. Please confirm your attendance.	0
URGENT: Your PayPal account has been locked. Click here to restore access: https://secure-paypal-reset.com	1

Column Descriptions:

• email_body: Text of the email
• label:
  • 1 = Phishing
  • 0 = Safe

🤖 Model Details (phishing_model.pkl)

• Type: scikit-learn Pipeline
• Steps:
  1. CountVectorizer
     • Converts email text into numerical feature vectors (Bag-of-Words approach).
     • Token pattern: Words with at least 2 letters.
     • Lowercase conversion enabled.
  2. Multinomial Naive Bayes Classifier
     • Classes: 0 = Safe, 1 = Phishing.
     • rained on labeled email datasets.
• Usage:
  • Loaded at runtime using joblib.load()
  • Takes processed text as input
  • Returns prediction as Phishing or Safe

🖼 Screenshots