Research On WEB Security Gateway Trojan Detection Optimization Technology

Currently people obtain and publish information mainly though web pages and their lives and works can no longer leave the Internet.However,while the internet provides people with information and convenience,it also brings a lot of security risks.Related research statistics show that virus as the most important threat in the network has been replaced by Trojans,and 90% of the host is infected by accessing the web page Trojans.Web Trojan has greater intensity than traditional Trojan in terms of the speed of spread,the scope of spread and the threats it makes.Therefore,how to detect the web Trojan before the Trojan intrusion and protected the users form the infringement of the Trojan has become the urgent problem of the network security.First of all,this paper proposes a web page flow analysis framework based on the minimum cache model aims at the time cost problem of the feature extraction at the web security gateway.Present researches mainly analyzes the web pages based on the DOM structure to obtain the characteristic information used for detection.The method needs to cache the entire page before it can be parsed.On the one hand,the timeliness of the analysis can not be guaranteed and on the other hand as the size of the webpage increases the parsing speed will reduce and the memory consumption will increase.This paper analyze the web pages by the flow analysis method in order to make sure the timeliness and the resolution speed will not be affected by the page size.And the method also ensure the minimum memory footprint while ensure the extraction of feature integrity though the minimum cache model.In this paper,the above two methods are used to compare the resolution time and memory occupation of web pages with different sizes,and verify the advantage of webpage parsed framework based on the minimum cache model in analyzing timeliness and reducing storage pressure.Secondly,in terms of the accuracy of WEB Trojan detection,this paper propose a self-encoder and BP neural network WEB Trojan detection technology.Current researches mainly uses the method of machine learning to detect the WEB Trojans.On the one hand there is a correlation between the extracted features in this paper and some classification methods do not suits these cases.On the other hand,the data in this paper comes from the WEB security gateway since there is less tagged data than normal data.If supervised learning is used alone for training,the accuracy of the test results will be flawed.In this paper,unsupervised learning based on stack denoising self-encoder is used to pre-train the data to obtain more robust feature information.Then minor adjust the nerual network structure through supervised training based on BP neural network.The method ensures that the detecting model gotten from training can still guarantee the accuracy of detection and detection efficiency even when the training data set has less tagged data.This paper verifies the advantages of WEB Trojan detection technology based on depth learning in both detection accuracy and detection efficiency by contrasting with other common classification methods.