Research On Phishing Website Detection Based On Data Mining Classification Algorithm

In recent years,with the development of the Internet,the Internet has gradually become a platform for people's activities such as work,study,shopping,and financial transactions.He has brought great convenience to people's work and life,and is prospering in the e-commerce and e-finance industries.At the same time,the issue of network security has become increasingly prominent.Among them,phishing websites are particularly harmful.The phishing website is a kind of social engineering-based attack method and is a criminal mechanism for criminals to deceive user identity data and financial accounts.The website,content and layout of phishing websites are very similar to those of real websites.This makes it difficult for netizens without security awareness to be deceived.Phishing brings huge economic losses to people and seriously hinders the development of e-commerce..Effectively curbing phishing websites is a guarantee of network security.Research on the detection and defense of phishing attacks is imminent.At present,there are achievements in the research of defense phishing websites at home and abroad.However,they all have flaws.In recent years,in the research of fishing website detection methods,the application of machine learning has made great achievements.The application of various classification algorithms in machine learning to phishing website detection and identification can effectively improve the detection efficiency.Therefore,this paper studies the phishing website identification method based on data mining classification algorithm.The focus of this paper is to analyze and compare the classification performance of different classifier algorithms in phishing website detection in machine learning.The main work of this paper is divided into three parts.First of all,in this paper,PCA principal component analysis method is used to analyze the importance of URL anomaly feature and multi-source fusion feature of selected phishing websites,and obtain effective features.Secondly,using three classical classification algorithms in machine learning—support vector machine,random forest,and RBF neural network—to generate classification models from web page feature sets,and conduct comparative experiments on them.Finally,aiming at the shortcomings of the three classification models,an improved new classifier model-KELM classification model is proposed based on this,and further evaluation and analysis are made based on URL anomaly characteristics and multi-source fusion characteristics through six evaluation indexes.The precision and performance of phishing website recognition.The experiment proves that it can effectively improve the recognition efficiency under the condition of using the KELM algorithm,and proves the effectiveness of the KELM classifier model in the detection of phishing websites.