The Research Of Web Security Detection Based On Hidden Markov Model And Convolutional Neural Network

According to the "Internet Security Threat Report" released by Symantec in 2017,network security is increasingly valued by the national security department,and web security is still an important issue in network security.With the rapid development of network technology and its applications,web applications have gradually become mainstream.Web application,as the main carrier of network applications,is widely used in various business systems.However,people often ignore the security of the web application.These vulnerable web applications are attacked by hackers and cause users'important data to be leaked,falsified,or destroyed.These web applications are vulnerable and can be easily attacked by hackers,causing users' important data to be leaked,falsified,or destroyed.So,It is crucial for protecting web applications to have an effective web attack detection method.This article focuses on web security detection technology based on hidden Markov models and convolutional neural networks.The main work is as follows:Traditional methods of detecting web attacks are detected by manually encoding attack features into corresponding rules.With the diversification of web attack methods,the shortcomings of traditional methods have become increasingly prominent.Aiming at the shortcomings of traditional detection technology,this paper proposes a web security detection technology method based on the bag of word model and hidden Markov model.This thesis first analyzes the basics of Web security theory and attack types.At the same time,the principle of Web attacks and corresponding defense measures are analysis in detail.In extracting payload features,this thesis summarizes the method of extracting payload features using N-grams model.As the feature extraction sliding window N increases,the number of generated subsequences grows exponentially and there are many repeated sequences.Based on this,we propose a bag of word model(BOW)and hidden Markov model to detect Web attacks.First,the word bag model is used to extract the payload feature.Then,the extracted features are encoded.Finally,the detection model is trained using a hidden Markov algorithm.Compared with the previous N-gram extraction feature algorithm experiment,the experimental results show that our experiment achieves higher detection rate under the condition of low false alarm rate.Based on the study of machine learning,aiming at the problems of slow web attack detection and inaccurate feature extraction,we propose to use convolutional neural network in deep learning to detect web attacks.Convolutional neural networks are developed based on the Keras machine learning library and the TensorFlow framework.This thesis uses the embedded word vector to learn to extract the semantic information of features.The convolutional neural network can automatically extract the effective features of the invasive samples.Finally,the detection model uses a classifier to identify samples.The detection model has a very high accuracy in the real production environment.