Phishing-URL-Detector

The Hybrid Phishing URL Detector is a web application designed to detect phishing URLs using a machine learning model

Comment

Phishing URL Detector

Project Screenshot
Screenshot of the project interface.

Table of Contents

Project Overview

The Hybrid Phishing URL Detector is a web application designed to detect phishing URLs using a machine learning model. It combines traditional URL features and keyword analysis to determine whether a given URL is malicious or safe. This tool helps users quickly check URLs to prevent phishing attacks and cyber fraud.

The project has a FastAPI backend for prediction and a frontend interface to interact with the model. Accuracy metrics are displayed on the frontend for transparency.

Features

  • Predict if a URL is safe or phishing.
  • Displays model accuracy and prediction confidence.
  • Interactive and responsive frontend.
  • Uses a hybrid dataset with features like URL length, number of dots, digits, HTTPS presence, and keywords.
  • Fast and efficient predictions using pre-trained ML model.
  • Can detect phishing attempts targeting Bangladesh domains (e.g., .bd) as well as global URLs.

Dataset

The dataset contains the following columns:

  • url – The URL string.
  • url_length – Length of the URL.
  • num_dots – Number of dots in the URL.
  • num_digits – Number of numeric characters.
  • has_https – Whether the URL uses HTTPS (1 = yes, 0 = no).
  • keyword_bank – Presence of phishing-related keywords.
  • local_keyword_flag – Flag for local domain keywords.
  • label – 0 for safe, 1 for phishing.

The dataset is preprocessed to train the ML model used in this project.

Technologies Used

  • Backend: FastAPI, Python 3.x
  • Frontend: HTML, CSS, JavaScript
  • Machine Learning: scikit-learn, pandas, numpy, scipy
  • Database: None (model-based detection)
  • Model: Pre-trained ML model (stored as model.pkl)
  • Tools: Joblib for model serialization

Installation

  1. Clone the repository:
https://github.com/ShoaibSikder/Phishing-URL-Detector.git
  1. Set up Python environment:
python -m venv venv
# Activate virtual environment
# On Linux/macOS
source venv/bin/activate
# On Windows
venv\Scripts\activate
  1. Install dependencies:
pip install -r backend/requirements.txt

Usage

  1. Run the FastAPI backend:
uvicorn app:app --reload --host 0.0.0.0 --port 8000
uvicorn backend.app:app --reload --port 8000

  1. Open the frontend:
    Open frontend/index.html in your browser.

  2. Test URLs:
    Enter any URL to check if it is safe or phishing. The frontend will display the prediction and model accuracy.

API Endpoints

  • POST /predict – Predicts if a URL is phishing or safe.
    Request body:
{
  "url": "https://google.com/login"
}

Response:

{
  "prediction": "Safe",
  "accuracy": 0.9
}

How it Works

  1. User inputs a URL in the frontend.
  2. The URL is sent to the FastAPI backend.
  3. Backend extracts features:
    • URL length
    • Number of dots
    • Number of digits
    • HTTPS presence
    • Keyword analysis
  4. Features are fed into the pre-trained ML model (model.pkl).
  5. Prediction result and model accuracy are returned and displayed.

Results & Accuracy

  • Model Accuracy: ~90.5%
  • Confusion matrix, ROC curve, and precision-recall analysis can be generated in the backend for evaluation.

License

This project is licensed under the MIT License.

Developed by Md. Shoaib Sikder.

Share this project:

Updates