In this paper, we provide a method for identifying malicious web pages using a crawling and classification approach that helps to support the automatic discovery of the malicious links. The proposed approach can successfully complete the crawling session even if the page requires partial page refreshment and authentication credentials. The evaluation of the proposed approach shows a higher accuracy compared to an existing approach with an overall accuracy of 72% in three custom applications. Moreover, the proposed approach will calculate the significance and the impact severances of each link on the website and it better differentiates malicious web pages and normal links