Detecting Phishing Websites using Machine Learning Algorithm

Abstract

In general, malicious websites aid the expansion of online criminal activity and stifle the growth of web service infrastructure. Therefore, there is a pressing need for a comprehensive strategy to discourage users from going to these sites online. We advocate for a method that uses machine learning to categories websites as either safe, spammy, or malicious. The proposed system is limited to examining the URL itself, rather than the contents of websites. As a result, it does away with both browser-based vulnerabilities and run-time delays. The proposed method outperforms blacklisting services in terms of generality and coverage since it makes use of learning techniques. There are three distinct categories for website addresses. Neutral Web sites provide average, risk-free functionality. For a website, “spam” refers to any attempt to overwhelm the user with advertisements or sites (such as false surveys and online dating sites). Malware is defined as a website designed by hackers to cause harm to computers and steal private data. The experimental data demonstrates a dramatic improvement in performance with the new model compared to the baseline.