摘要
Webshell is a web backdoor which can be used to remotely control web servers by hackers. Due to the continuous development of escaping technology, Webshell detection has become more and more difficult. Based on the analysis of evolving Webshell detection technology in recent years, we proposed Multi-classifier ensemble Model in this paper. Firstly, effective information of PHP files are extracted including static characters, grammar character and corresponding opcode; secondly, different base classifiers and modified classifier are trained and analyzed for further model, at last the ensemble model based on stacking is proposed and verified. Our dataset collected from multiple GitHub open source projects. The mothod proposed in this paper could ultimately achieve the accuracy of 98.447% and precision of 99.227%, which showed excellent performance compared with mainstream detection tools.