An anti-phishing method based on feature analysis

Abstract

Since the rapid advancement in computer networks and ebusiness technologies, massive numbers of sales transactions are performed on the World Wide Web on daily basis. These transactions necessitate online financial payments and the use of ebanking hence attracting phishers to target online users' credentials to access their financial information. Phishing involves developing forged websites that are visually identical to truthful websites in order to deceive online users. Different anti-phishing techniques have been proposed to reduce the risks of phishing mainly by educating users or using automated software. One of the main challenge for automated anti-phishing tools is to determine the more influential features in order to detect phishing activities. This article addresses this problem by conducting a thorough analysis using filtering methods against real phishing websites data. The methodology employed is based on data mining method called RIPPER to determine the performance of the classification systems derived by RIPPER and according to different evaluation measures such as error rate, false positives and false negatives.