An Efficient Phishing Detection Framework Based on Hybrid Machine Learning Models

Abstract

The paper discusses an improved phishing detection system that uses hybrid machine learning for analyzing URL-based features. Further improvement of prior work using better feature selection as well as ensemble methods is discussed. The proposed model enhances classification accuracy, precision, and recall by utilizing an innovative hybrid ensemble approach that integrates Logistic Regression, Support Vector Machine, and Decision Tree, alongside newly incorporated evaluation techniques. The results manifest high improvement in performance metrics as compared to the previous methods. Various comparisons and benchmarks with other methods have further proved the robustness of our proposed system for detecting malicious URLs.