🛡️ Phishing Website Detection

🚀 Project: End-to-End Modular ML Pipeline for Detecting Phishing Websites

This project builds a scalable and modular machine learning pipeline to detect phishing websites based on URL and network-based features. It uses a production-grade architecture where each ML lifecycle stage is isolated and implemented as a separate component, enabling reuse, testing, and seamless CI/CD deployment to the cloud.

Dagshub

---

🔍 Problem Statement

Phishing websites are a common cyber threat. This ML system is trained on over 10,000+ website entries and 30 extracted features to predict if a website is legitimate or phishing.

---

🧱 Project Architecture

• Data Ingestion
• Data Validation
• Data Transformation
• Model Training
• Model Evaluation
• Model Pusher
• API Deployment (FastAPI)
• CI/CD via GitHub Actions to AWS EC2

---

🛠️ Tech Stack

Tool	Purpose
Python	Core programming language
Pandas/Numpy	Data preprocessing
Scikit-learn	ML models and evaluation metrics
FastAPI	Serve model as REST API
Docker	Containerization
GitHub Actions	CI/CD pipeline for build/test/deploy automation
AWS EC2 & ECR	Hosting and container registry
MLflow	Experiment tracking
Dagshub	Remote tracking (optional)

---

⚙️ Core Libraries

python-dotenv, pandas, numpy, scikit-learn, dill, pyyaml, fastapi,
uvicorn, pymongo, mlflow, dagshub, certifi, setuptools, scipy

---

🧠 Dataset Overview

• Rows: 10,000+
• Target: Result (1 = phishing, -1 = legitimate)
• Columns: 30 engineered network/URL features

Feature Examples	Type
having_IP_Address	Numeric
URL_Length	Numeric
SSLfinal_State	Numeric
Page_Rank	Numeric
Abnormal_URL	Numeric
Google_Index	Numeric

---

🔁 CI/CD Pipeline (GitHub Actions)

• CI Stage: Code linting, Unit testing
• CD Stage: Docker image build ➝ Push to ECR ➝ Deploy on EC2 via self-hosted runner
• Auto Deploy: Every push to main triggers full deployment

---

🌐 API Access

Once deployed, the FastAPI app is accessible at:

http://<EC2-Public-IP>:8080/docs

Use /predict endpoint to make POST requests with website features. Use /train endpoint to run the pipeline.

---

📁 Folder Structure (Simplified)

📦 MLOPS-Network-Security-System
├── app.py
├── main.py
├── Dockerfile
├── requirements.txt
├── setup.py
├── README.md
├── .github/
│   └── workflows/
│       └── main.yml
├── networksecurity/
│   ├── __init__.py
│   ├── cloud/
│   │   └── s3_syncer.py
│   ├── components/
│   │   ├── data_ingestion.py
│   │   ├── data_validation.py
│   │   ├── data_transformation.py
│   │   ├── model_trainer.py
│   ├── constant/
│   ├── entity/
│   │   ├── config_entity.py
│   │   └── artifact_entity.py
│   ├── exception/
│   ├── logging/
│   ├── pipeline/
│   │   ├── training_pipeline.py
│   │   └── batch_prediction.py
│   └── utils/
│       ├── file_utils/
│       └── ml_utils/
├── data_schema/
├── prediction_output/
├── final_model/
├── notebooks/
├── templates/
├── valid_data/
├── temp_network_data/
├── Artifacts/
│   ├── data_ingestion/
│   ├── data_validation/
│   ├── data_transformation/
│   └── model_trainer/

---

📦 Deployment Notes

• Project is containerized with Docker
• ECR is used as container registry
• GitHub Actions runner on EC2 handles deployment
• ML model tracked with MLflow (local or remote)

---

👨‍💻 Author

Vikramaditya — Final Year IT Student passionate about Computers and Programming. This project is a part of my ML studies and course by Krish Naik Sir.

---