Phishing Detection Using Machine Learning: Performance Evaluation of Classification Algorithms
Year: 2025
Type: article
Abstract: Phishing attacks pose a major cybersecurity threat by exploiting human vulnerabilities through deceptive websites and emails. Detecting and classifying these phishing attacks is crucial in identifying essential patterns in the attacks. However, such effort is extensive and thus requires support from machine learning-based techniques. This study evaluates the effectiveness of seven machine learning (ML) models—Decision Tree (DT), K-Nearest Neighbors (KNN), Naïve Bayes (NB), Random Forest (RF), Support Vector Machine (SVM), Artificial Neural Network (ANN), and Logistic Regression (LR)—on two phishing datasets. The models used 10-fold cross-validation for evaluation, with the performance metrics evaluated as accuracy, precision, recall, and F1-score. RF and ANN performed the best on the Phishing Website Dataset with the highest accuracies of 96.63% and 96.54%, respectively. On the Website Phishing Dataset, DT and RF achieved the highest accuracy of 87.82%. NB exhibited the poorest performance, attributable to insufficient recollection. Confusion matrices and heatmaps validated that RF, ANN, and SVM yield the most dependable classification outcomes. The results indicate that ensemble learning (RF) and deep learning (ANN) are exceptionally effective for phishing detection. Subsequent efforts should concentrate on hyperparameter optimization, feature selection, and hybrid modeling to improve accuracy.
Authors: Mohd Arfian Ismail, Hilal H. Nuha, Setyorini, Suhaila Mohd. Yasin
Language: English
Topic: Spam and Phishing Detection
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4416059221