Phishing URL Detection in Isolated Systems

As phishing attacks continue to rise in frequency, scale, and impact, detecting these attacks becomes even more critical. Applying modern machine learning algorithms to URL data allows for the reliable detection of phishing websites. These predictions can be used to protect end users, train security tools, and identify malicious actors before they can victimize innocent users. In this paper, thirty-four machine-learning models are trained on sample data, and five models are selected and tested against a test dataset. The overall best-performing classifiers are identified, and statistical results are provided. These models accurately identify phishing websites in 92.45% of cases using only publicly available URL and domain information.