ADVANCED HYBRID FEATURE SELECTION AND BAYESIAN OPTIMIZATION FOR PHISHING URL DETECTION

Abstract

Phishing attacks are a major cybersecurity threat that aim to steal sensitive information by redirecting users to fraudulent websites. Traditional blacklist- and rule-based methods often remain insufficientespecially against zero-day attacks. This study presents a systematic methodological benchmark framework for phishing URL detectionintegrating hybrid feature selectionBayesian hyperparameter optimizationmodel comparisonensemble learningexplainability analysisand additional validation strategies. Experiments were conducted on a dataset of 88647 URLs with 111 featuresevaluating 7 hybrid feature selection methods3 Bayesian optimization techniquesand 252 optimized model configurationscomplemented by deep learning baselines. The findings show that hybrid feature selection and optimization provide a clear contributionparticularly for tree-based models. The LightGBM model optimized on the L1+Boruta feature set with Scikit-Optimize produced the strongest single-model performance (Test Accuracy: 97.28%F1: 95.99%AUC: 99.54%)while the best ensemble (CAT+XGB+LGBMSoft Voting) reached 97.34% accuracy96.17% F1and 99.60% AUC — a comparable level of performance. Deep learning models yielded lower performancewhile additional validation experiments demonstrated that the ensemble structure provided a consistent improvement over the baseline across different data splits.