Pregledni rad
Machine Learning Approaches for Phishing Detection Based on URL Analysis
Ivana Hartman Tolić
orcid.org/0000-0002-6805-1328
; Fakultet elektrotehnike, računarstva i informacijskih tehnologija Osijek, Kneza Trpimira 2b, 31000 Osijek, Hrvatska
*
Mirta Vujnovac
; III. gimnazija Osijek, Kamila Firingera 14, 31000 Osijek, Hrvatska
* Dopisni autor.
Sažetak
Phishing attacks have posed a significant threat to cybersecurity in recent years. Phishing is a form of social engineering in which attackers provide misleading information via fake websites in order to trick the victim into disclosing private information to obtain further information or gain a financial advantage. With the rapid development of technology and phishing tactics, access to information and the frequent exchange of information, effective methods for detecting fake URLs are needed. The goal is to evaluate the effectiveness of different models in classifying malicious and legitimate web addresses without analyzing the content of the page. This study aimed to evaluate the effectiveness of various machine learning and deep learning models in classifying malicious and legitimate web addresses without analyzing page content. Experimental results show that convolutional neural networks (CNNs) can achieve accuracy rates of up to 98.7%, while ensemble models such as Random Forest and XGBoost also demonstrate high accuracy, exceeding 96%, significantly outperforming traditional approaches like logistic regression.
As phishing strategies continue to evolve, adaptive models such as ensemble learning techniques, deep learning architectures will be fundamental to securing online security and crucial to understanding how to effectively counter emerging cybersecurity threats.
Ključne riječi
Cyber attacks, Ensemble models, SMOTE, Social engineering, URL classification
Hrčak ID:
334234
URI
Datum izdavanja:
30.7.2025.
Posjeta: 467 *