A minimum enclosing ball-based support vector machine approach for detection of phishing websites

In this paper, a novel approach based on minimum enclosing ball support vector machine (BVM) to phishing Website detection is proposed, which aims at achieving high speed and high accuracy for detecting phishing Website. In order to enhance the integrity of the feature vectors, we first perform an analysis of the topology structure of website according to the DOM tree and use the Web crawler to extract 12 topological features of the website. Then, the feature vectors are detected by BVM classifier. Compared with the general SVM, this method has relatively high precision of detecting, and complements the disadvantage of slow speed of convergence on large-scale data. The experimental results show that the proposed method has better performance than SVM, and further validate the validity and correctness of our scheme.