Classification of Features for detecting Phishing Web Sites based on Machine Learning Techniques

Abstract

Phishing is one of the most common and dangerous attacks among cybercrimes. The aim of this attack is to hack the user information by accessing the credentials that is used by individuals and any of the organizations. Phishing websites contents and web-based information contains various hints. The victim’s confidential data is expected by the phishing sites by deriving them to surf a phishing website that resembles to legitimate website, which is one of the criminal attacks prevailing in the internet. Phishing websites is similar to cyber threat that is targeting to get all the credential-based information such as information accessed from the credit cards and social security numbers. Till now there is no specific solution that can detect phishing attacks and also truly unpredictable which includes numerous components and also criteria that are not stable. The purpose of this project is to perform Extreme Learning Machine (ELM) based classification for 30 features including Phishing Websites Data in UC Irvine Machine Learning Repository database. There are different types of features based on web pages. Hence, to prevent phishing attacks we must use a specific web page feature. We proposed a model based on machine learning techniques like Naïve Bayes to detect phishing web pages. For results assessment, ELM was compared with other machine learning methods such as Naïve Bayes (NB), ANN and detected to have the highest accuracy of 89.3%.