Malicious Website Detection Optimization in Enterprise Cybersecurity Management: A Machine Learning-Based Decision Support Approach

The increasing proliferation of malicious web platforms presents critical risks to organizational cybersecurity, necessitating the creation of advanced detection systems that adapt to evolving cyber threats. Conventional identification approaches frequently prove inadequate against sophisticated attack patterns. This paper proposes an innovative website threat analysis framework combining structured feature selection with t-SNE visualization in data-centric methodologies. Through targeted feature optimization, we improve model transparency and processing effectiveness while minimizing noise in complex datasets. The analytical framework evaluates five widely-adopted classification algorithms, XGBoost, KNN, Naive Bayes, Random Forest, and SVM, to assess detection capabilities. Validation tests using the Malicious/Benign Websites dataset reveal that Random Forest and XGBoost achieved superior performance in identifying malicious domains, maintaining high accuracy and balanced F1-scores. These outcomes highlight the value of embedding adaptive machine learning solutions into corporate security infrastructures to optimize threat recognition, operational responsiveness, and system resource allocation. The study offers practical guidance for enhancing cybersecurity budgeting decisions and developing preemptive protection frameworks within enterprise risk management contexts.