A Prior-based Transfer Learning Method for the Phishing Detection

In this paper, we introduce a prior-based transfer learning method for our statistical machine learning classifier which based on the logistic regression to detect the phishing sites that relies on our selected features of the URLs. Because of the mismatched distributions of the features in different phishing domains, we employ multiple models for different regions. Since it is impossible for us to collect enough data from a new region to rebuild the detection model, we adjust the existing models by the transfer learning algorithm to solve these problems. The proposed algorithm was evaluated on a real-world task of detecting the phishing websites. After a number of experiments, our proposed transfer learning algorithm achieves more than 97% accuracy. The result demonstrates the use of this algorithm in the anti-phishing scenario is feasible and ready for our large scale detection engine.