Research On Network Intrusion Detection Algorithm Based On Machine Learning

Under the background of continuous development of science and technology in recent years,more and more high and new technologies have been applied to the computer network,bringing great convenience to people's production and life.Meanwhile,all kinds of our private information are continuously uploaded to the network,including our property information,identity card information,location information,website browsing information and so on,which are all completely saved in the network.More and more criminals illegally obtain people's information through network attacks.Traditional network security measures can no longer meet people's demand for information security.Along with the continuous development of machine learning technology,machine learning is becoming more and more powerful in extracting data and information,and its classification performance for various problems is continuously improving.Applying machine learning technology to intrusion detection systems has become a research topic for more and more scientists.The essence of intrusion detection problem is classification problem.There are many classification methods in machine learning,including convolutional neural network,Naive Bayesian,SVM,etc.These algorithms can achieve high classification accuracy on balanced data sets.However,the intrusion detection data set is unbalanced data set,and the number of samples for network attacks is far less than that for normal networks.The classification accuracy of the above algorithm on unbalanced data sets will be reduced,even the samples of minority classes cannot be identified,and the training speed is slow.In order to solve these problems,this paper proposes to apply convolution neural network algorithm and random forest algorithm in machine learning to intrusion detection model,which can effectively improve the accuracy of the overall classification of the model and the recognition rate of minority classes.The main work of this paper includes the following aspects:Firstly,in terms of the problems of slow training speed and low detection rate existing in the current network intrusion detection model,this paper proposes a network intrusion detection model based on convolution neural network.Through increasing the number of convolution layers and pooling layers in the model,the model can fully extract the high-dimensional information in the data and improve the detection rate of network data.Meanwhile,the resampling algorithm based on RANDOM?SMOTE is used to reduce the imbalance rate of data sets and further improve the detection rate of the model on CICIDS2017 data sets.Finally,a feature extraction method based on feature importance and correlation analysis is used in the model to make preliminary feature selection on CICIDS2017 data set to reduce the amount of input data,thus reducing the training time of the model.Secondly,in terms of the low detection rate of the current network intrusion detection model for minority classes in unbalanced data sets,this paper proposes a network intrusion detection model based on random forest algorithm.The random forest algorithm has better classification performance on unbalanced data sets.The random forest algorithm is combined with the improved SMOTE algorithm(K-SMOTE algorithm)to improve the detection rate of the model for minority classes.Finally,a network intrusion detection model based on convolution neural network and random forest is proposed.The model uses convolution neural network to divide normal class and attack class,and then uses random forest algorithm to divide 14 attack classes.Finally,compared with the experimental results in other papers,it can be found that the overall classification accuracy of the model in this paper is higher,and the recognition rate for minority classes is also higher.