Detecting Phishing Websites Using Machine Learning

Abstract

The entire world is digitizing at a rapid pace. However, the ever-evolving transformation comes with its fair share of vulnerabilities, opening the doors wider for cybercriminals. One of the common types of attacks criminals indulge in these days is phishing. It involves creation of websites to dupe unsuspecting users into thinking that they are on a legitimate site, making them disclose confidential information like bank account details, usernames, and passwords. This paper aims to find out accurately if a URL is reliable or not, in other words, if it is a phished one or otherwise. Machine Learning (ML) based models provide an efficient way to detect these phishing attacks. This research paper focuses on using three different ML algorithms—Logistic Regression, Support Vector Machine (SVM), and Random Forest Classifier in order to find the most accurate model to predict whether a given URL is safe or not. To achieve this, the respective models are trained using a pre-existing data set and then tested as to whether they can accurately classify the websites or not. The algorithms are also compared based on performance measures like Precision, Accuracy, F1 Score, and Recall to deduce which one of the three is most efficient and reliable for classification and prediction.