International Journal of Advanced Research in Computer and Communication Engineering

Abstract: Phishing has become one of the most pervasive and damaging forms of cybercrime, targeting unsuspecting internet users through fraudulent websites that mimic legitimate ones. These malicious platforms deceive users into revealing sensitive credentials such as passwords, financial information, and personal identification details. The continuous evolution of phishing tactics—such as sophisticated URL obfuscation, dynamic content manipulation, and social engineering—renders traditional detection mechanisms increasingly ineffective. Conventional defense strategies, including blacklists, heuristic filters, and rule-based approaches, fail to detect newly emerging or “zero-day” phishing websites that are not yet cataloged in known databases. Hence, there is an urgent need for an adaptive, intelligent, and automated solution that can accurately detect phishing websites in real time without dependence on third-party services.
This study focuses on developing a machine learning-based phishing website detection system that leverages URL-based, domain-based, and HTML content-based features to distinguish between legitimate and phishing websites. The core idea is to train classification models capable of learning behavioral patterns and structural differences inherent in phishing websites. The dataset used for experimentation consists of over 60,000 URLs, equally divided into phishing and legitimate samples, collected from verified and publicly available repositories. Feature extraction plays a pivotal role in this system. Important features include URL length, use of special symbols, number of subdomains, domain registration age, SSL certificate presence, hyperlink patterns, and term frequency–inverse document frequency (TF-IDF) vectors derived from the website’s HTML content.

Downloads: | DOI: 10.17148/IJARCCE.2025.141192