Enhanced Classification Method for Homograph Attack Detection

Internationalized Domain Name (IDN) homograph is a web security attack in which the attackers deceive the computer users about what websites they are accessing by using homologous domain names. Recently, the growth of IDN homograph attack has become a severe problem with a significant probability of criminality like frauds for web users. This paper proposes an enhanced classification method for IDN homograph detection by utilizing the Structural Similarity Index (SSIM). Compared to the existing approach, the experiment results showed that our improved classification method could increase the accuracy from 95.07% to 96.18% and decrease the false positive rate from 3.92% to 3.23%. Moreover, we apply a multi-group-of-classifier method to our model, which can further increase the accuracy of 98.34% with a false positive rate of 3.77%. We also conducted an empirical analysis of the IDN homograph data and the SSIM classification approach's training processes to discuss why our method outperforms the existing method in homograph detection.