Phishing Detection System Through Hybrid Machine Learning Based on URL

Abstract

ABSTRACT
Currently, numerous types of cybercrime are organized through the internet, with phishing
attacks being a primary focus of this study. Despite its origins in 1996, phishing has evolved
into one of the most severe and dangerous forms of cybercrime on the internet. Phishing
typically involves email distortion to create deceptive communications, followed by the use of
fraudulent websites to extract sensitive information from unsuspecting individuals. While
various studies have addressed the prevention, identification, and awareness of phishing
attacks, a comprehensive and effective solution to thwart them remains elusive. This is where
machine learning becomes crucial in defending against phishing attacks. The proposed study
utilizes a phishing URL-based dataset from a well-known repository, comprising attributes of
both phishing and legitimate URLs collected from over 11,000 websites. After preprocessing
the data, several machine learning algorithms were applied and designed to detect and prevent
phishing URLs, thereby protecting users. The machine learning models employed in this study
include decision tree (DT), linear regression (LR), random forest (RF), naive Bayes (NB),
gradient boosting classifier (GBM), K-neighbors classifier (KNN), support vector classifier
(SVC), and a proposed hybrid LSD model. The LSD model combines logistic regression,
support vector machine, and decision tree (LR+SVC+DT) with both soft and hard voting
mechanisms to enhance accuracy and efficiency in defending against phishing attacks. To
optimize the performance of the proposed LSD model, the canopy feature selection technique,
cross-fold validation, and Grid Search Hyperparameter Optimization techniques were used.
The effectiveness of the models was evaluated using various metrics, including precision,
accuracy, recall, F1-score, and specificity. Comparative analysis of the results shows that the
proposed approach outperforms other models, delivering superior results in terms of both
accuracy and efficiency.