ADVANCED MACHINE LEARNING APPROACH FOR PHISHING WEBSITE IDENTIFICATION USING META ENSEMBLE METHODS

Abstract

Phishing attacks are a major cybersecurity threat that affects millions of internet users and organizations
worldwide and can lead to severe financial and privacy losses if not detected early. Traditional phishing
detection methods rely heavily on blacklist databases and manual rule-based analysis of website features,
which can be time-consuming and often fail to identify newly created phishing websites. Recent
advancements in artificial intelligence and machine learning have enabled automated phishing detection
systems that assist in identifying malicious websites more efficiently. This study proposes an advanced
meta-ensemble machine learning framework for phishing website identification using multiple
classification models. The framework combines the predictive capabilities of algorithms such as Random
Forest, Support Vector Machine, Gradient Boosting, and XGBoost with a meta-classifier to generate more
accurate final predictions. The ensemble architecture integrates multiple feature types including URLbased,
domain-based, and content-based attributes to improve the system’s understanding of phishing
patterns. In addition, ensemble learning techniques such as stacking and voting are used to enhance model
robustness and reduce prediction errors. Experimental results demonstrate improved detection accuracy
and better reliability compared to traditional single machine learning models. The proposed framework
can assist cybersecurity systems in making faster and more accurate decisions for detecting phishing
websites and protecting users from online fraud.