Malicious Uniform Resource Locator Detection Using Wolf Optimization Algorithm and Random Forest Classifier
No Thumbnail Available
Date
2021
Journal Title
Journal ISSN
Volume Title
Publisher
Machine Learning and Data Mining for Emerging Trend in Cyber Dynamics: Theories and Applications
Abstract
Within the multitude of security challenges facing the online community,
malicious websites play a critical role in today’s cybersecurity threats. Malicious
URLs can be delivered to users via emails, textmessages, pop-ups or advertisements.
To recognize these malicious websites, blacklisting services have been created by
the web security community. This method has been proven to be inefficient. This
chapter proposed meta-heuristic optimization method for malicious URLs detection
based on genetic algorithm (GA) and wolf optimization algorithm (WOA). Support
vector machine (SVM) as well as random forest (RF) were used for classification of
phishingweb pages. Experimental results showthatWOAreduced model complexity
with comparable classification results without feature subset selection. RF classifier
outperforms SVM based on the evaluation conducted. RF model without feature
selection produced accuracy and ROC of 0.972 and 0.993, respectively, while RF
model that is based onWOA optimization algorithm produced accuracy of 0.944 and
ROC of 0.987. Hence, in view of the experiments conducted using two well-known
phishing datasets, this research shows that WOA can produce promising results for
phishing URLs detection task.
Description
Keywords
Phishing detection, Meta-heuristic, Genetic algorithm, Wolf optimization, Machine learning