Phishing Prediction on Website Updates with Novel Features Through Machine Learning

Phishing attacks are becoming increasingly sophisticated, making it difficult to predict them using traditional security measures. One technique that has gained traction is leveraging machine learning algorithms to identify phishing websites. This research study presents a machine learning rooted approach to possibly predict phishing attempts by thoroughly analyzing website updates. The proposed approach involves training a XGBoost classifier using a dataset of legitimate and phishing websites. For this approach, created a new dataset labelled as “phish_predict” that includes novel features to enhance the accuracy of the classifier in identifying changes in web pages that could indicate a phishing attempt. This study's goal was to identify phishing websites using ML techniques. Three well-known algorithms, K-NN, SVM, and XGBoost, were tested, and their effectiveness was analyzed using a range of measures, involving the performance metrics. Among these algorithms, XGBoost had the greatest accuracy of 100% in predicting phishing websites. This novel approach makes an effective addition to traditional security measures for predicting phishing attempts.