The Classification and Detection of Malicious URLs Using a Machine Learning Based Approach

Abstract

AbstractThe world wide web has become the most cardinal and salient source of information, for which in today's technological era, web pages are incessantly being utilized to broadcast pernicious activity over the web. Hence, the perpetual technological advancements have enthralled cybercriminals to easily exploit the web environment by embedding malicious code on these web pages which permits oblivious and innocent users into becoming victims by visiting these harmful pages with a click of a button. Such detrimental and malicious web pages are created by cyber-attackers to promote and host viruses and exploit frauds, attacks, and scams.Consequently, within an organization, emails containing Uniform Resource Locators (URLs) cause a considerable amount of risk towards an organization as these links become instrumental in terms of giving systematic control to the cyber-attackers. There is a critical need of filtering employees’ incoming emails in accordance with the maliciousness of the URL to protect the web environment users from such threats. Accordingly, the premier objective of this proposed paper is to constructively classify and detect malicious URLs while utilizing a machine learning based approach. This will lead towards sufficiently decreasing the number of URL-based attacks that may target organizations via email, protecting the users of the web from phishing, scam, defacement, and malware web attacks. Hence, the conceptualization behind the proposed AI framework is to construct a multi-classification model to effectively thwart and protect organizations from detrimental and malicious web-based attacks by classifying a URL as either a defacement, benign, malware, spam, or a phishing-based URL. The general framework for the proposed malicious URL detection system is illustratively displayed in Figure 1.