An optimal machine learning-based algorithm for detecting phishing attacks using URL information

Abstract

In recent years, more websites have been collecting personal information for many processes, such as banks, internet connections, and government services. The public needs to provide all personal information, such as Aadhar, PAN, date of birth, and phone number. The personal and sensitive information is at risk of being used for phishing attacks through URL manipulation. In addition, a phishing attack cause’s financial and reputational loss. Hence protecting sensitive information by adapting required protection is extremely valuable for global security. To overcome this, we proposed a method to detect phishing attacks based on previous history, including the duration of operation, customer reviews, web traffic, and the URL. Based on these parameters, the proposed optimal machine learning-based algorithm (OmLA) analyze the previous information about URLs and predict whether it is phishing- or legitimate. As per simulation and performance analysis, the proposed method outperforms conventional methods such as random forest (RF), support vector machine (SVM), and genetic algorithms (GA) by 8%, 18%, and 23%, respectively in terms of accuracy. Additionally, it achieves detection times of 0.2%, 0.6%, and 0.9%, respectively, and excels in response times of 0.45%, 0.56%, and 0.62%, respectively.