Comparison of Adaboost with MultiBoosting for Phishing Website Detection

Developments in the Internet introduced new trends. Innovations in Internet bring challenges to our security and expectations of privacy. The anonymous structure of the Internet has such a big attack surface that possess increased risks from cyber-attacks. Attackers introduced novel techniques, such as phishing, to take their vulnerable information such as account IDs, usernames, passwords, etc. Phishing is defined as mimicking a creditable company’s website aiming to take private information of a user. It is a challenge to realize whether a web page is legitimate or phishing, because of its semantics-based attack structure that mostly exploits the Internet users’ vulnerabilities. Several remedies proposed in order to prevent phishing. Although software companies introduced novel anti-phishing techniques that utilize heuristics, blacklists, visual and machine learning methods, only one single magic bullet cannot eliminate these threats entirely. Ensemble machine learning techniques are promising methods employed to distinguish phishing attacks. In this study, an intelligent phishing website detection framework is presented. We employed different machine learning models to classify websites as legitimate or phishing. Several classification techniques were employed in order to implement precise intelligent phishing website detection framework. Classification accuracy, F-measure and area under receiver operating characteristic (ROC) curves (AUC) are utilized to evaluate the performance of the machine learning methods. Experimental results revealed that Adaboost with SVM has outperformed best among the classification methods by achieving the highest accuracy 97.61%.