A Research On SQL Injection And XSS Attack Detection Based On Deep Learning

The advent of new Internet technologies and artificial intelligence technologies,great changes in traditional production and lifestyle have taken place.The network environment is becoming more and more complicated,and the network security situation is becoming increasingly severe.Almost all network applications are facing various kinds of security threats and network attacks.The weakness of traditional network security detection technology is increasingly obvious.There is an urgent need of new types of network attack detection technology that is adapted to the current network environment.Therefore,combining deep learning with network attack detection technology has a significant reality Significance and value.URLs are used to launch various of attacks frequently,and SQL injection and XSS account for 80 percent of all type of attacks caused by URLs.In this work,we take advantage of deep learning to propose a system to detect SQL injection attack and XSS attack in URLs.Firstly,based on the analysis of existing URL attack detection technologies,a method for URL feature representations was designed.This method is capable of normalizing and vectorizing URL requests by sematic analysis and the vector obtained by this feature representation method can retain most useful information with removing redundant information.Secondly,we proposed a method to combine multiple deep learning models.All previous works used single deep learning model to detect attacks and faced the problem of being cheated and attacked.In order to get a stable system,we proposed a method to combine multiple deep learning models,which can improve the detection accuracy and stability of the system.Thirdly,this work designs and implements a complete Web attack detection system,and tests each module of the system on multiple open data sets and real data sets.Besides,we compared the proposed system with some existing systems and the experimental results fully prove the feasibility and effectiveness of the detection technology proposed in this work.Finally,in order to test the system in actual scenarios,we deployed a distributed detection system in a real network environment and used a variety of network security tools to conduct real-time attacks against the target system.Experimental results,include 100%attack detection rate,show that the detection system has an excellent ability to detect web attacks in the actual network environment.