Optimized Decision Forest for Website Phishing Detection

The development of web and internet technology has resulted in its application in a wide range of services. This has resulted in an increase in the number of cybersecurity issues over the years, the most famous of which is the phishing attack, in which hostile websites impersonate genuine websites t...

Full description

Main Authors: Balogun, A.O., Mojeed, H.A., Adewole, K.S., Akintola, A.G., Salihu, S.A., Bajeh, A.O., Jimoh, R.G.
Format: Article
Institution: Universiti Teknologi Petronas
Record Id / ISBN-0: utp-eprints.29355 /
Published: Springer Science and Business Media Deutschland GmbH 2021
Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85120689064&doi=10.1007%2f978-3-030-90321-3_47&partnerID=40&md5=456b754af9aa5d92230c7532745d536a
http://eprints.utp.edu.my/29355/
Tags: Add Tag
No Tags, Be the first to tag this record!
Summary: The development of web and internet technology has resulted in its application in a wide range of services. This has resulted in an increase in the number of cybersecurity issues over the years, the most famous of which is the phishing attack, in which hostile websites impersonate genuine websites to acquire naïve users� data required for illegal access. Current mitigation measures, including anti-phishing software and machine learning (ML) approach, have proven to be successful in identifying phishing operations. Hackers, on the other hand, are coming up with new techniques to get around these counter-measures. Nonetheless, given the dynamism of phishing efforts, there is a constant requirement for novel and efficient website phishing detection solutions. In this study, an optimized decision forest (ODF) method for detecting website phishing is proposed ODF involves the use of a genetic algorithm (GA) for the selection of optimal diverse individual trees in a forest to generate an efficient sub-forest. Specifically, accurate and diverse trees from a decision forest are passed into GA as an initial population to generate a more robust forest with high efficacy. The performance of the proposed ODF is evaluated using three phishing datasets from the UCI repository. Findings from the experimental results revealed that ODF performed better than selected baseline classifiers. Particularly, ODF recorded a high detection accuracy (98.37), AUC (0.999), f-measure (0.98), MCC (0.967) values with a low false-positive rate (0.016). In addition, ODF outperformed some existing ML-based phishing attack models. Consequently, the proposed ODF method is recommended for dealing with sophisticated phishing attacks. © 2021, The Author(s), under exclusive license to Springer Nature Switzerland AG.