🔐 Phishing URL Detection System | Network Security Project

An end-to-end Machine Learning system to detect phishing websites based on over 30 URL-based and HTML-based features. This project demonstrates a complete MLOps workflow from data processing to model deployment.

---

🚀 Project Highlights

• ✅ 30+ Engineered Features extracted from URLs to detect phishing behavior
• ✅ 98% Accuracy achieved using Random Forest (tuned) on labeled dataset
• ✅ Fully modular ETL pipeline: Data Ingestion → Validation → Transformation → Training
• ✅ Web app deployment with FastAPI + Streamlit on AWS EC2
• ✅ CI/CD with GitHub Actions, Docker, MLflow, DVC, and S3 for artifact tracking
• ✅ MongoDB Atlas used for structured feature and logging storage
• ✅ Seamless feature extraction, API serving, and browser-based testing

---

🧠 Tech Stack

Layer	Tools/Technologies
Language	Python
ML Libraries	scikit-learn, pandas, numpy
Deployment	FastAPI, Streamlit, Docker, AWS EC2
MLOps Tools	MLflow, DVC, GitHub Actions, AWS S3
Database	MongoDB Atlas
CI/CD	GitHub Actions + Docker

---

🗂️ Project Structure

---

⚙️ Pipeline Overview

1. Data Ingestion

• Reads raw data from CSV or S3
• Stores raw artifact using DVC

2. Data Validation

• Checks schema consistency
• Validates missing/null/unexpected values

3. Data Transformation

• Extracts 30+ handcrafted features
• Handles outliers, encodes categories

4. Model Training

• Trains multiple ML models (RF, DT, LR, XGBoost)
• Evaluates via accuracy, precision, recall
• Logs experiments using MLflow

5. Prediction & Serving

• FastAPI server serves predictions
• Feature extraction on-the-fly from user-input URL
• Streamlit provides interactive web interface

---

🖥️ Running the Project Locally

🔧

git clone https://github.com/yourusername/phishing-detection.git
cd phishing-detection



python -m venv myenv
source myvenv/bin/activate  # On Windows: mvenv\Scripts\activate
pip install -r requirements.txt

uvicorn app:app --reload
streamlit run streamlit_app.py
docker build -t phishing-detector .
docker run -p 8000:8000 phishing-detector

🧪 Example URLs for Testing
Type	Example URLs
✅ Good	https://www.youtube.com
https://www.nytimes.com
❌ Phishing	http://198.51.100.23/login
http://paypal-login.com

🙌 Acknowledgements
Dataset: Phishing Websites Dataset - UCI ML Repo

Inspired by real-world phishing detection techniques and OWASP recommendations

📬 Contact
Ali Abdullah
📧 aliabdullah@example.com
🔗 LinkedIn • GitHub