Research On Network Intrusion Detection Based On Deep Learning

With the development of network technology,the risk of intrusion is also increasing,so intrusion detection is more and more important.Since it was put forward in 1980 s,it has gone through many stages of technological development.From the beginning of audit technology,to machine learning used in the field of intrusion detection,to the development and application of in-depth learning in recent years,intrusion detection technology has been a big step forward.Different from the traditional machine learning algorithm,the traditional machine learning algorithm will decline with the increase of data volume,while the deep neural network needs more data,a large number of data samples,the better the training effect for the deep neural network.Because the existing intrusion detection data generally exist the problem of data imbalance,although deep neural network has the characteristics of high performance,the problem of data imbalance also partially limits the performance of deep neural network.Based on this,this paper improves the application of deep learning in intrusion detection field by constructing new intrusion detection model and improving data set.The main work of this paper is as follows:(1)Aiming at the low detection rate and high false alarm rate caused by data imbalance and feature redundancy in the training process,a deep learning model(TDDL)based on secondary decision is proposed,In the first stage,the automatic encoder is used for feature processing,and in the second stage,the neural network is used for classification,so as to reduce the feature redundancy and balance the deviation of normal data,so as to improve the detection effect.(2)Improved convolutional neural network(CNN)model for intrusion detection.There are two main improvements.One is to improve the network model so that it can better match the intrusion detection data set.The other is to optimize CNN's pooling layer by using the mixed random pooling method.(3)Aiming at the problem of data imbalance of intrusion detection model,the data level optimization is adopted.In the process of training,it improves the way of sending data into neural network model.In this paper,KDD99 data set is used for the experiment,and CNN and BP neural network are mainly used as the basis for the comparative experiment.The experiment proves that the network model proposed in this paper is better,which can improve the accuracy and reduce the false alarm rate at the same time.