Abstract:

In recent years, as malicious websites harm to every aspect of the user, the detection of malicious web site URL is becoming increasingly important. At present, the detection of malicious URL mainly includes black and white list technology and machine learning classification algorithm.However, the black and white list technology can do nothing while the URL is not in list. And each machine learning classification algorithm has some data which it is not good at. In this paper, we propose a malicious URL multi-level filtering detection model. By training the threshold of each layer filter, the filter can directly determine the URL when it reaches the threshold. Otherwise, the filter leave the URL to next layer. Therefore, every classifier can deal with the data it is good at, this paper uses an example to verify that the model can improve the accuracy of URL detection.

Key words: malicious URL, black and white list technology, machine learning, multi-layer filtering model