Application of Machine Learning for Real-Time Phishing Attack Detection

Over the years, the Internet has been exploited to carry out a range of cyber attacks, withphishing being the most prominent one. Increasingly sophisticated techniques of phishinghave threatened the security of many Internet-based systems. To be able to detect suspiciouswebsites is a potential first step in reducing the amount of phishing attacks occurring daily.This paper outlines the development and implementation of a platform to detect phishingwebsites. It highlights the pressing need for early detection of possible phishing attacks toprevent data theft, frauds, etc. The system uses machine learning algorithms to distinguishlegitimate websites from phishing websites and generate a prediction to be used for theplatform. A user interface is implemented to have two parts. The first part includes a textfield for entering a URL, which the ML model processes to give a prediction that getsdisplayed to the user. Another module gathers URLs as they arrive from an API and scansthem for potentially suspicious websites. The final ML model, a Random Forest classifierwith 27 estimators, had an accuracy of 96.12% and F1 score of 95.94%. Futureenhancements and research directions are also discussed for further development of thesystem.