Abstract:

With multi-source hyperlink indices, this study proposes a malicious websites identification model based on hyperlink analysis and classification rule. The performance and associative rules of four types of classification rule in identifying malicious websites are analyzed by comparing with four typical machine learning classifiers. By analyzing the extracted rules, the hyperlink indices from Alexa and Moz play an important role for the malicious website identification. Compared with four typical machine learning classifiers, the proposed identification model not only extracts a group of identification rules for malicious websites, but also has better performance in identifying malicious websites. This study can not only expand the use of hyperlink analysis in the area of malicious websites identification, but also build an efficient model and extract easy-to-understand rules in identifying malicious websites.

Key words: Malicious website, Hyperlink analysis, Classification rules, Machine learning, Website identification