Research On Intrusion Detection Technology Based On Deep Neural Network With Complex Structure

In recent years,with the rapid development of the Internet and society,the network has greatly changed people's learning,work and lifestyle.But at the same time,the security problems caused by network security are becoming more and more serious.Due to the increasing types of attacks,the increasing complexity of attack methods and the”low threshold” of making network attacks,the traditional network intrusion detection technologies are facing the following problems:(1)The data source of network traffic is highly heterogeneous and the scale is increasing rapidly;(2)The performance of classical machine learning algorithms depends on Feature Engineering;(3)New types of attacks are emerging one after another;(4)Some of the existing automated network intrusion detection devices have high false positives.It is difficult for the current network intrusion detection technology to solve the above problems effectively.This paper proposes a supervised deep learning algorithm in network intrusion detection technology by using deep learning and word embedding methods,which can effectively capture complex spatio-temporal features in network traffic data.This method can improve the accuracy and reduce the false alarm rate at the same time.We finally complete the word embedding-based algorithm for network intrusion detection and the construction of deep learning model for intrusion detection.The main research work in this paper is as follows:(1)This paper compares the advantages and disadvantages of convolution neural network and recurrent neural network in network intrusion detection.It is instructive to construct intrusion detection model based on deep neural network with complex structure.(2)This paper proposes a preprocessing method of network traffic data based on word embedding.Word embedding technology is widely used in the field of natural language processing.Word vectors obtained by word embedding technology not only have lower dimensions than one-hot coded vectors,but also maintain semantic relevance.In this paper,word embedding technology is introduced to extract features of the application layer load and reduce the dimension.Some head features are combined to obtain low-dimensional and high-quality feature vectors.(3)This paper proposes a deep neural network with complex structure to learn the features of network traffic.In this paper,convolutional neural network is used to learn the spatial features of traffic packets,and recurrent neural network is used to learn the time series features of multiple packets in a session.Aiming at the current problems of traditional network intrusion detection technology,this paper applies deep learning and word embedding method to network intrusion detection technology through sufficient literature research and discussion.This paper compares the advantages and disadvantages of convolution neural network and recurrent neural network,and proposes an effective method to capture the spatio-temporal characteristics of network traffic data.The supervised deep learning algorithm,which is based on word embedding and in-depth learning,completes the construction of deep neural network model with complex structure and is used in network intrusion detection.The experimental results show that the model has high accuracy and low false alarm rate,and has made a certain contribution to deploy the deep-learning-based network intrusion detection system in the real world.It has certain theoretical and practical significance.