Phishing URL Detection Using Supervised Machine Learning Algorithms

Abstract

Phishing is one of the security threats that has been in existence for the past decades. It is the process of tricking unsuspecting users with the aid of baits such as URL (Uniform Resource Locator) links to lure victims from a secured platform to an unsecure platform with the intent of stealing valuable information such as credit card details, bank token, and identity etc. Many researchers have developed several methods to combat these threats although been effective, it has failed to combat the increasing adaptive ability of the phishers who attack unsuspecting victims in recent times.  In this study therefore, we propose an adaptive data collection system for use in the automatic detection of phishing URL. An adaptive data collection system for legitimate and phishing URL based on certain selected features will be used in the extraction of features using a structured feature extraction vector model which will be trained and tested using five (5) classification models built from Random Forest, Decision Tree, XGboost, Adaboost and KNN. The results obtained from the five models show that model from Random Forest outperforms the other models from other classifiers with an accuracy of 99.38%, precision of 90.21% and a recall rate of 82.55%. The model developed from the random forest classifier was then selected and deployed on a phishing detection platform developed. In the future we seek to expand the features and also seek to explore the usage of deep neural network as the features keeps expanding.