A Webshell Detection Algorithms via Multi-parameter Comprehensive Evaluation Based on Text Recognition

Abstract

Webshell refers to the malicious backdoor that hackers can obtain the execution authority of arbitrary commands on the network servers. However, Webshell uses a variety of encryption and obfuscation technologies, which increase the difficulty of Webshell detection. To solve the current problems with Webshell detection, We propose a multi-parameter comprehensive evaluation Webshell detection algorithm based on text recognition, called Webshell-TR, and build a Webshell detection system WebShellKing according to this algorithm. The innovation of our method is to build a multi-parameter comprehensive evaluation detection algorithm by integrating seven parameters of Webshell detection, and improve the Webshell detection ability through the fusion judgment of multiple parameters. In order to verify the proposed algorithm, we build a benchmark data set WebSHBDB for Webshell detection, and carry out a controlled experiment. The experimental results show that WebShellKing has better recall rate than other popular Webshell detection tools, and can identify more covert Webshells.