A Study On Webshell Detection Model Based On Machine Learning

With the continuous development of the Internet,the connection between websites and people's lives has become close.As a result,the security of websites has become more and more significant.According to the "2017 China Internet Cyber Security Report" released by the China National Internet Site Center,29,236 websites were implanted in the back door of Webshell in 2017,including 1,339 government websites.So the research on the detection of Webshell attack is meaningful and necessary.This paper first introduces the basic concepts of Webshell and some conventional traditional Webshell detection methods,analyzes the performance of these methods,and clarifies that in the background of the increasingly diverse Webshell attack methods,new methods are needed to implement Webshell detection.The thesis research found that scholars in related fields have applied machine learning and deep learning algorithms in Webshell detection.Their models have made great progress in performance compared with traditional methods,but there is still room for improvement.The method of text preprocessing is introduced later in this article.The collected webpage text file samples are pre-compiled to obtain an intermediate operation code,and then the word sequence is converted into a purely digital feature vector using a word vector conversion algorithm to obtain an input set of the detection algorithm model.In this paper,the built-in multi-layer neural network model is used to verify the experimental results,and a variety of other machine learning related algorithms are used to verify the performance differences of different algorithms.After comparison,the best detection model algorithm is found.In the end,this paper provides an implement of the Webshell detection system.