🎣 Phishing Email Detector

A machine learning-powered web application that detects phishing emails using Natural Language Processing and Logistic Regression. Built with Python, Streamlit, and scikit-learn.

🚀 Features

• Real-time Detection: Instantly classify emails as Safe or Phishing
• High Accuracy: Trained on 175,000+ email samples
• Web Interface: User-friendly Streamlit application
• Confidence Scoring: Shows prediction confidence percentages
• Modular Design: Separate scripts for data processing, training, and prediction

📊 Project Overview

This project implements a complete machine learning pipeline for phishing email detection:

1. Data Processing: Clean and preprocess email dataset
2. Feature Engineering: TF-IDF vectorization for text analysis
3. Model Training: Logistic Regression classifier
4. Web Deployment: Interactive Streamlit application
5. Standalone Prediction: Command-line prediction script

🛠️ Technology Stack

• Python3
• Machine Learning: scikit-learn, pandas, numpy
• Web Framework: Streamlit
• Text Processing: TF-IDF Vectorization
• Model Persistence: joblib

📁 Project Structure

Phishing Detector/
├── app.py                      # Streamlit web application
├── data_checker.py             # Data exploration script
├── data_cleaning.py            # Data preprocessing
├── model_train.py              # Model training script
├── predict.py                  # Standalone prediction
├── Phishing_Email.csv          # Original dataset
├── cleaned_phishing_emails.csv # Processed dataset
├── AI_Models/
│   ├── phishing_model.joblib   # Trained model
│   └── vectorizer.joblib       # TF-IDF vectorizer
└── README.md                   # Project documentation

🚀 Quick Start

Prerequisites

pip install streamlit pandas scikit-learn joblib numpy

Installation

1. Clone or download the project

   git clone <repository-url>
   cd "Phishing Detector"

2. Install dependencies

   pip install -r requirements.txt

Usage

Option 1: Web Application (Recommended)

1. Run the Streamlit app

   streamlit run app.py

2. Open your browser to http://localhost:8501

3. Paste email content and click "Check Email"

Option 2: Command Line Prediction

python predict.py

🔧 Model Training (Optional)

If you want to retrain the model:

1. Check the dataset

   python data_checker.py

2. Clean the data

   python data_cleaning.py

3. Train the model

   python model_train.py

📈 Model Performance

• Algorithm: Logistic Regression
• Features: 7,000 TF-IDF features
• Training Data: ~140,000 emails (80% split)
• Test Data: ~35,000 emails (20% split)
• Evaluation: Accuracy, Precision, Recall, F1-Score

☕ Support the Project

If this project helped you or you found it useful, consider buying me a coffee! Your support helps maintain and improve this project.

Other ways to support:

• ⭐ Star this repository
• 🍴 Fork and contribute
• 📢 Share with others
• 🐛 Report issues and bugs

🎯 How It Works

1. Text Preprocessing

• Removes stop words (common English words)
• Converts text to lowercase
• Handles missing data

2. Feature Extraction

• TF-IDF Vectorization: Converts email text to numerical features
• Term Frequency: How often words appear in each email
• Inverse Document Frequency: How rare words are across all emails

3. Classification

• Logistic Regression: Binary classifier (Safe vs Phishing)
• Probability Scores: Confidence percentages for predictions

4. Prediction Pipeline

Email Text → TF-IDF Transform → Model Prediction → Result + Confidence

📝 Example Usage

Web Interface

1. Paste suspicious email content
2. Click "Check Email"
3. Get instant result with confidence score

Sample Inputs

Phishing Email Example:

Congratulations! You've won $1000! Click here immediately to claim your prize!

Safe Email Example:

Hi John, just confirming our meeting tomorrow at 2 PM. Please let me know if this works.

🔍 Key Features Explained

TF-IDF Vectorization

• Purpose: Convert text to numbers for machine learning
• Max Features: 7,000 most important words
• Stop Words: Removes common words like "the", "and", "is"

Logistic Regression

• Why: Fast, interpretable, good for text classification
• Output: Binary classification with probability scores
• Training: 1000 max iterations for convergence

Model Persistence

• Saved Files: Both model and vectorizer saved separately
• Consistency: Ensures same preprocessing for new predictions

🚨 Security Considerations

• Input validation for email content
• No storage of user email data
• Secure model file handling
• Error handling for corrupted files

🔧 Suggested Customizations

Modify TF-IDF Parameters

# In model_train.py
vectorizer = TfidfVectorizer(
    stop_words='english',
    max_features=7000,  # Adjust number of features
    ngram_range=(1, 2)  # Add bigrams
)

Change Model Algorithm

# Replace LogisticRegression with other algorithms
from sklearn.ensemble import RandomForestClassifier
model = RandomForestClassifier(n_estimators=100)

📊 Performance Monitoring

The model provides:

• Prediction confidence (probability scores)
• Classification reports during training
• Accuracy metrics on test data

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Test thoroughly
5. Submit a pull request

📄 License

This project is open source and available under the MIT License.

🆘 Troubleshooting

Common Issues

1. "Model files not found" error

   • Run python model_train.py first to create model files

2. Import errors

   • Install missing packages: pip install package_name

3. Dataset not found

   • Ensure Phishing_Email.csv is in the project directory

4. Streamlit not opening

   • Check if port 8501 is available
   • Try: streamlit run app.py --server.port 8502

📧 Contact

For questions or support, please open an issue in the repository.

---

⚠️ Disclaimer: This tool is for educational and research purposes. Always use multiple layers of security for email protection in production environments.