🔒 Network Security – Phishing Detection System

A production-grade machine learning system for detecting phishing websites using 30 security features.

✨ Features • 🏗️ Architecture • 🚀 Quick Start • 🔬 How It Works • 📚 Examples

---

🎯 Overview

Network Security is a comprehensive ML-based phishing detection system that analyzes websites using 30 engineered features to identify potential phishing threats.

It includes:

• A FastAPI backend for real-time and batch predictions
• A Streamlit frontend for interactive analysis
• A full ML pipeline for training, validation, and evaluation
• MLflow integration for experiment tracking and model registry
• Dockerized deployment for production use

From this:

Upload a CSV with website features

To this:

Prediction Results
------------------
Website 1 → Legitimate
Website 2 → Phishing
Website 3 → Suspicious

precision: 0.9890399837629389
recall_score: 0.9926665308616827
f1_score: 0.9908499389995933

⏱️ Prediction time: milliseconds per sample ✅ Status: Production-ready ML system

---

💡 Why I Built This

The Problem Phishing websites continue to grow rapidly, and manual inspection is slow, inconsistent, and error-prone.

Goal Build an end-to-end machine learning system that can:

• Automatically detect phishing websites
• Support real-time and batch predictions
• Track experiments and models professionally
• Be deployed easily using Docker

What I Learned

• Designing modular ML pipelines improves maintainability
• MLflow simplifies experiment tracking and versioning
• FastAPI is ideal for low-latency ML APIs
• Streamlit accelerates building data apps
• Docker makes ML systems reproducible and portable

---

✨ Key Features

🎯 Core Functionality

• Batch prediction using CSV uploads
• Manual feature input for single-website analysis
• Real-time phishing detection
• RESTful FastAPI backend
• Interactive Streamlit dashboard

📊 Analysis Capabilities

• 30 website features analyzed
• Risk assessment scoring
• Accuracy, Precision, Recall, F1-Score
• Feature-wise breakdown
• Exportable CSV results

🔧 ML Pipeline

• Automated data ingestion from MongoDB
• Data validation and transformation
• Hyperparameter tuning using GridSearchCV
• Multiple models evaluated
• MLflow experiment tracking
• Model versioning and registry

---

🏗️ Architecture

High-Level Flow

User Input / CSV
        │
        ▼
 Streamlit UI
        │
        ▼
 FastAPI Backend
        │
        ▼
 Preprocessing
        │
        ▼
 Trained ML Model
        │
        ▼
 Predictions + Metrics

Project Structure

Network_Security/
├── app.py                      # FastAPI backend
├── streamlit_app.py            # Streamlit frontend
├── main.py                     # Training pipeline
├── Dockerfile                  # Docker configuration
├── requirements.txt            # Dependencies
├── setup.py                    # Package setup
├── .env                        # Environment variables
│
├── networksecurity/
│   ├── components/
│   │   ├── data_ingestion.py
│   │   ├── data_validation.py
│   │   ├── data_transformation.py
│   │   └── model_trainer.py
│   │
│   ├── entity/
│   │   ├── config_entity.py
│   │   └── artifact_entity.py
│   │
│   ├── pipeline/
│   │   ├── training_pipeline.py
│   │   └── batch_prediction.py
│   │
│   ├── utils/
│   ├── exception/
│   ├── logging/
│   └── constant/
│
├── final_model/
├── Artifacts/
├── Network_data/
├── prediction_output/
└── templates/

---

🛠️ Tech Stack

Component	Technology
Backend	FastAPI, Uvicorn
Frontend	Streamlit
ML	scikit-learn, NumPy, Pandas
MLOps	MLflow, DagsHub
Database	MongoDB, PyMongo
Visualization	Plotly, Matplotlib
Deployment	Docker, Docker Compose

---

🚀 Quick Start

Prerequisites

• Python 3.10+
• MongoDB (local or Atlas)
• Git

Installation

git clone <repository-url>
cd Network_Security

python -m venv myenv
source myenv/bin/activate  # Linux/Mac
myenv\Scripts\activate     # Windows

pip install -r requirements.txt
pip install -e .

Configuration

Create a .env file:

MONGODB_URL_KEY=mongodb+srv://username:password@cluster.mongodb.net/?appName=Cluster0

MLflow settings (in model_trainer.py):

os.environ["MLFLOW_TRACKING_URI"] = "https://dagshub.com/your-username/Network-Security.mlflow"
os.environ["MLFLOW_TRACKING_USERNAME"] = "your-username"
os.environ["MLFLOW_TRACKING_PASSWORD"] = "your-token"

---

🔬 How It Works

Phase 1: Data Ingestion

• Load phishing dataset from MongoDB
• Store raw data in artifacts

Phase 2: Data Validation

• Check schema
• Handle missing values
• Validate feature ranges

Phase 3: Data Transformation

• Feature preprocessing
• Train-test split
• Scaling and encoding

Phase 4: Model Training

• Train multiple models
• Perform GridSearchCV
• Select best model
• Log metrics to MLflow

Phase 5: Prediction

• Load trained model
• Apply preprocessing
• Generate predictions
• Return risk score

---

📚 Example Usage

Start Backend

python app.py

Docs: http://localhost:8000/docs

Start Frontend

streamlit run streamlit_app.py

UI: http://localhost:8501

---

🌐 API Endpoints

GET /

Redirects to docs

GET /train

Triggers model training

curl -X GET http://localhost:8000/train

POST /predict

Batch prediction

curl -X POST http://localhost:8000/predict \
  -F "file=@data.csv"

---

🎓 Model Training

Models Evaluated

• Random Forest
• Decision Tree
• Gradient Boosting
• Logistic Regression
• AdaBoost

Feature Values

• -1 → phishing
• 0 → suspicious
• 1 → legitimate

---

🐳 Docker Deployment

docker build -t networksecurity:latest .
docker run -p 8000:8000 --env-file .env networksecurity:latest

Docker Compose:

version: '3.8'
services:
  api:
    build: .
    ports:
      - "8000:8000"
    env_file:
      - .env

---

📊 Model Performance

• Accuracy: ~95.2%
• Precision: ~94.8%
• Recall: ~93.5%
• F1-Score: ~94.1%

---

🤝 Contributing

1. Fork the repository
2. Create feature branch
3. Commit changes
4. Push and open PR

---

📄 License

MIT License © 2025 Vivek Kumar Gupta

---

👨‍💻 Author

Vivek Kumar Gupta AI Engineering Student | MLOps & ML Systems

GitHub: @vivek34561 LinkedIn: vivek-gupta-0400452b6

---

🙏 Acknowledgments

• UCI Machine Learning Repository
• MLflow
• FastAPI
• Streamlit

---

Made with ❤️ for Cybersecurity

_{Built to explore end-to-end machine learning systems}