Naive and Neighbour Approach for Phishing Detection

Phishing is a cybercrime that involves tricking the user into entering sensitive information often by disguising as an authentic website. It has become very critical for us to devise solutions in order to protect user information. Our proposed work uses machine learning models for classification of websites to prevent frequent security attacks. We utilise various non-text based features from authentic sources. The inexistence of text features in our dataset makes the process of detection easier for Domain Registrars and Search Engines. Results indicate that KNN outperforms other classifiers by giving an accuracy of 95% in detection of phishing websites. However, Naive Bayes does exceptionally well in identifying the authentic websites.