The Phishing Detection Method Based On Fuzzy Associative Classification

With the development of the Internet,more and more Internet crime began to appear.Phishing is one of the most common form of crime.Phishing sites imitate the famous sites through technical means to steal user information and financial accounts so that users lose money.Phishing have a great harm to the development of Internet and property safety of Internet users.So detection methods of phishing have important practical significance.This paper studies the phishing detection method that proposed by the scholars,combines with feature selection and fuzzy associative classification.Phishing features for the current phishing detection method extracted contains a large number of irrelevant and redundant features.For this problem this paper proposes a feature selection algorithm combined LVF algorithm with CFS algorithm--LVCFS algorithm.We extract sensitive URL and Web page feature from the test page.Because the LVF algorithm only considers the correlation between features and category,without considering the association between features and features,and therefore LVF algorithm can't effectively remove redundant features.There will be some redundant features,using CFS algorithm to remove redundant features.LVCFS algorithm combines the advantages of LVF algorithm and the CFS algorithm can effectively remove irrelevant and redundant features.For the phishing site detection,the existing detection methods have less accurate.When Phishing features contain continuous numeric features,there aren't better detection methods.We propose an algorithm based on fuzzy associative classification named CFARWV.CFARWV is obtained by improving the fuzzy associative classification algorithm CFAR.CFARWV use the weighted vote FRM method instead of CFAR 's single winner FRM method,and the use of improved weights instead of confidence as rule weights.Using the algorithm LVCFS to select sensitive feature,then using CFARWV classification algorithm trained CFARWV classifier,after training use the trained classifier to classification to arrive at a determination result.Finally,this paper experiment to verify the proposed LVCFS algorithm,then compare with LVF,CFS,ReliefF algorithms.CFARWV classification algorithm have experimental verification and compare with Ripper,CMAR,CPAR,CFAR algorithms.This paper compare with the current phishing detection and look forward to the future work.