<p>Phishing detection remains a critical concern in the domain of cybersecurity, as such attacks potentially result in substantial financial losses and reputational damage to affected organizations. Conventional methodologies, including rule-based systems and fundamental machine learning classifiers, frequently demonstrate inadequate efficacy in providing comprehensive protection against evolving phishing threats. To address these limitations, this research proposes a novel integration of Multi-Objective Optimization (MOO) algorithms with Extreme Gradient Boosting (XGBoost). The proposed methodology implements evolutionary strategies for optimal feature subset selection while concurrently utilizing gradient boosting mechanisms for classification tasks, thus optimizing multiple performance metrics simultaneously. Our model incorporates a feature relevance-guided population initialization strategy and adaptive threshold mechanisms to balance model complexity with predictive power. Experimental results demonstrate that the proposed MOO/XGBoost model consistently outperforms other techniques. Additionally, the computational efficiency of the model, with a complexity of <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="10586_2025_5390_Article_IEq1.gif" Format="GIF" Height="23" Rendition="HTML" Resolution="72" Type="Linedraw" Width="153" /> </InlineMediaObject> <EquationSource Format="TEX">$O\left( {n^{2} {\text{log}}n + m \times n^{2} } \right)$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>O</mi> <mfenced close=")" open="("> <mrow> <msup> <mi>n</mi> <mn>2</mn> </msup> <mtext>log</mtext> <mi>n</mi> <mo>+</mo> <mi>m</mi> <mo>×</mo> <msup> <mi>n</mi> <mn>2</mn> </msup> </mrow> </mfenced> </mrow> </math></EquationSource> </InlineEquation>, makes it suitable for large and high-dimensional datasets. These findings highlight the potential of MOO/XGBoost as a promising approach for phishing detection and an effective feature selection strategy to enhance cybersecurity defenses against modern threats. </p>

A novel multi-objective optimization- XGBoost based feature selection and optimization for enhanced phishing website detection

期刊论文
发表日期：

2025年9月3日

查看全文

Santosh Kumar Birthriya,
Priyanka Ahlawat,
Ankit Kumar Jain

摘要

Phishing detection remains a critical concern in the domain of cybersecurity, as such attacks potentially result in substantial financial losses and reputational damage to affected organizations. Conventional methodologies, including rule-based systems and fundamental machine learning classifiers, frequently demonstrate inadequate efficacy in providing comprehensive protection against evolving phishing threats. To address these limitations, this research proposes a novel integration of Multi-Objective Optimization (MOO) algorithms with Extreme Gradient Boosting (XGBoost). The proposed methodology implements evolutionary strategies for optimal feature subset selection while concurrently utilizing gradient boosting mechanisms for classification tasks, thus optimizing multiple performance metrics simultaneously. Our model incorporates a feature relevance-guided population initialization strategy and adaptive threshold mechanisms to balance model complexity with predictive power. Experimental results demonstrate that the proposed MOO/XGBoost model consistently outperforms other techniques. Additionally, the computational efficiency of the model, with a complexity of $O\left( {n^{2} {\text{log}}n + m \times n^{2} } \right)$ $O (n^{2} log n + m \times n^{2})$ , makes it suitable for large and high-dimensional datasets. These findings highlight the potential of MOO/XGBoost as a promising approach for phishing detection and an effective feature selection strategy to enhance cybersecurity defenses against modern threats.