Detecting and Classifying Phishing Websites by Machine Learning

By imitating legitimate websites, phishing websites aim at stealing information from people. This paper thoroughly depicts a method of detecting phishing websites with joint features. Original data was crawled form PhishTank. SVM and bayes methods are mixed in training classifier. Wireshark is used in the process of packet flow detection. After going through training, the classifier can detect 1000 website per second. A data set of 21615 phishing and legitimate websites is used in the comparative study. In addition, 51 features are used to train and test the classifiers.