PERFORMANCE ANALYSIS OF SELECTED MACHINE LEARNING ALGORITHMS IN THE DETECTION OF PHISHING ATTACKS ON VULNERABLE WEBSITES

Abstract

Phishing is a cyber-attack where attackers pretend to be trustworthy, to trick individuals into providing sensitive information. Phishing also involves directing web users to fake websites that closely resemble legitimate ones and asking their victims to enter their personal information. It is imperative to restrict all forms of phishing websites or URLs. This study analyzes the performance of several machine-learning algorithms in creating models capable of detecting phishing websites. To achieve this central goal, massive phishing website detector datasets were retrieved from an online open repository, Kaggle. Relevant libraries in Python were explored for pre-processing, uploading, and partitioning of the datasets for training and testing the model. Models were created based on each of the six machine-learning algorithms implemented in this study. Evaluation of each model reveals that; Random Forest has the highest accuracy of 96.7%. followed by Support Vector Machine which records 96.4%. However, evaluating the model created using the Naive Bayes Classifier shows the lowest accuracy of 60.5%. This study has revealed the strength of each algorithm for detecting phishing websites; it also unveiled the procedures that need to be taken to mitigate and curb the spread of phishing attacks.