Sign in Subscribe
Machine learning

Effective Phishing Detection System Developed by University of Baghdad Researchers

Strategic Insights

— 1 min read

Researchers at the University of Baghdad have developed an effective two-phase phishing detection system, employing unsupervised feature selection and supervised classification. The system utilizes machine learning to precisely identify malicious websites, addressing a critical problem in cybersecurity. The development aims to enhance the accuracy and efficiency of phishing detection, with promising results showing an accuracy of 99% using the voting ensemble technique with feature selection.

Key Takeaways:

  • The system is designed to address the critical problem of phishing detection in cybersecurity by utilizing machine learning with an efficient feature selection method.
  • The two-phase system employs unsupervised feature selection and supervised classification, which improves the accuracy and efficiency of phishing detection.
  • The Genetic algorithm is used in the first phase to identify the best set of features, while the K-means clustering algorithm is utilized to divide the dataset into groups with similar traits.
  • In the second phase, the Genetic algorithm is used to enhance the classification process by identifying the best set of features in each group.
  • A voting ensemble technique is applied, combining Support Vector Machine (SVM), Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Adaptive boosting (AdaBoost) models.
  • The system achieves an accuracy of 99% using the voting ensemble technique with feature selection compared to 77.3% without feature selection.
  • The model performance experiences a significant boost through the GA-optimised feature selection, reducing computational complexity and improving key metrics such as accuracy, precision, and F1-score.
  • The system demonstrates the positive impact of K-Means clustering in improving classification accuracy for specific data groups.
  • The research concludes that integrating feature selection with ensemble learning is effective for phishing detection, with the scalability and efficiency of such a solution demonstrated in real-world applications.

Statistics:

  • Accuracy: 99% (with feature selection) and 77.3% (without feature selection)
  • Number of URLs: 11,430
  • Features: 87
  • Accuracy improvement: 21.7% (using voting ensemble technique with feature selection compared to without feature selection)
  • F1-score improvement: 26.9% (using voting ensemble technique with feature selection compared to without feature selection)

Sources:

  • "K-Gen PhishGuard: an Ensemble Approach for Phishing Detection with K-Means and Genetic Algorithm" (Al-Khawarizmi Engineering Journal, 2025,21(2))
  • Al-Khwarizmi College of Engineering - University of Baghdad (Publisher)
  • https://doi.org/10.22153/kej.2025.04.011 (free version of the journal article)

Read more