Auto Malicious Websites Classification Based on Naive Bayes Classifier

Abstract

In recent years, the spread of malicious websites has had an increasingly serious impact on people's lives. To solve this problem, in this paper, we mainly focus on achieving the factor analysis of the website category and accurate identification of unknown information, so as to classify the benign and others, which help users avoid the risk of malicious websites. In the procedure, Naïve Bayes and other effective methods are used to calculate probability to test and train the model of website classification. Our result shows the response of benign and other tests well compares to others, suggesting that the Naïve Bayes method is more suitable to solve the differentiation of good websites, which accuracy can be reached to 90 percent. Besides, the training model of that classification is more accurate in that datasheet. Furthermore, if the data is more abundant and the technical bottleneck can be solved, the realization of highest accuracy of website classification using Naïve Bayes is possible.