Paper Title: Feature Selection Framework for Optimizing ML-based Malicious URL Detection

Overview

This repository contains the implementation and resources for our research paper on optimizing machine learning-based malicious URL detection through intelligent feature selection. Malicious URLs are responsible for billions of dollars in annual losses through cyber-attacks, including spam, phishing, social engineering, and malware distribution.

Our approach addresses a critical challenge in ML-based URL detection: selecting the most effective representation features without extensive manual feature engineering or domain expertise.

Key Contributions

• Hybrid Feature Selection Framework: Combines multiple methods (Information Gain, Genetic Algorithms, Random Forest) to identify optimal feature subsets
• Efficient Training: Reduces computational resources and training time while maintaining detection accuracy
• Automated Feature Engineering: Minimizes the need for manual feature selection and domain expertise
• Comparable Performance: Achieves performance on par with traditional approaches using significantly fewer features

Citation

If you use this work in your research, please cite:

@INPROCEEDINGS{10778786,
  author={Shah, Sajjad H. and Garu, Amit and Nguyen, Duong N. and Borowczak, Mike},
  booktitle={2024 Cyber Awareness and Research Symposium (CARS)}, 
  title={Feature Selection Framework for Optimizing ML-based Malicious URL Detection}, 
  year={2024},
  volume={},
  number={},
  pages={1-6},
  keywords={Training;Uniform resource locators;Machine learning algorithms;Accuracy;Computational modeling;Simulation;Feature extraction;Vectors;Random forests;Genetic algorithms;Malicious URL detection;cybersecurity;machine learning;feature selection},
  doi={10.1109/CARS61786.2024.10778786}}

Paper

Read the full paper: IEEE Xplore

License

This project is licensed under the MIT License - see the LICENSE file for details.