Identifying gambling and porn websites with image recognition

Abstract

Gambling and porn websites are more and more harmful to the health and growth of the youth with the rapid development of the Internet, however, the text contents and URLs based website classification methods could not get satisfying on gambling and porn websites detection because domain names of them change fast. Meanwhile, the visual based website classification has gotten perfect results in phishing website detection which encourages us. Therefore, we introduce the visual feature to identify gambling websites and porn websites in this paper. Firstly, we develop a website screenshot tool which could save the full contents of a website to be a image, Secondly, the effective feature is chosen by BoW model to recognize the screenshots of gambling websites and porn websites, and the appropriate parameters are chosen to promote the efficiency of classification. Finally, experimental results on our collected gambling websites and porn website datasets demonstrate that our proposed method is able to recognize the gambling and porn websites and gets satisfying results.