A Machine Learning Technique to Detect Phony Websites

Phishing websites are a threat to internet usersas they mimic the appearance of authentic websites to acquiresensitive data. Traditional phishing attacks can be preventedwith techniques like blacklist methods, rule-based methods,and similarity study. Still, the methods remain inadequate dueto various disadvantages like poor scaling potential, inability totackle zero-day attacks, and higher computationalcomplexities. Therefore, in the current manuscript, detection ofthe authenticity of websites through a supervised machinelearning technique has been discussed. Based on features likeURL length, hostname length, appearance of IP address,presence of specific data like IP address in URL, appearance ofspecial characters, digit ratios in the URL and domain name,domain age, Yahoo indexing status, and Google ranking status,a supervised machine learning technique has been employed.An adequate amount of data has been accumulated from datapools like Phish Tank and Kaggle. Various machine learningclassifiers such as Logistic Regression, Decision Tree, SupportMachine, and Random Forest have been employed for training.Moreover, the experimentally obtained results show the highestefficiency in terms of accuracy, precision, recall, and F1-scoreusing the Random Forest classifier. Thus, the system efficientlyidentifies both existing as well as new types of phishingwebsites.