Malicious code detection method and system based on recurrent neural network

Abstract

The invention relates to a malicious code detection method and system based on a recurrent neural network, and the method comprises the steps: obtaining a to-be-detected target code, carrying out the preprocessing of the target code, and obtaining a target code operation code sequence; performing vectorization processing on the operation code sequence of the target code to correspondingly obtain a word vector of the target code; and inputting the word vector of the target code into a pre-trained LSTM malicious code detection model for target detection, and detecting whether a malicious code exists in the target code or not. Compared with the traditional machine learning algorithm, the method has the advantages that the accuracy is effectively improved, the loss rate is reduced, the manual operation is reduced, and the detection capability of unknown Webshell is also effectively enhanced.

Description

technical field [0001] The invention relates to the field of network security, in particular to a malicious code detection method and system based on a cyclic neural network. Background technique [0002] Webshell is a web page backdoor program, also known as a Trojan horse, and is a web management tool that can operate web servers. Hackers often use various loopholes to upload their own written Webshell to the Web server to manage files, databases, etc. Hackers can perform a series of malicious operations such as uploading and downloading files, invoking related system commands, etc. on the website embedded in the Webshell, and what's more, they can bypass WAF through various obfuscation methods. Webshells are extremely stealthy and difficult to detect, and their variety has been greatly expanded over time. According to the type of script, it can be divided into PHP Trojan, ASP Trojan, JSP Trojan and CGI Trojan; according to its function, it can be divided into big horse,...

AI Technical Summary

Problems solved by technology

However, traditional machine learning often requires manual adjustment of the rule set. The error is large and the transferability is poor, and a lot of manpower and material resources are wasted.

Images