Research On Network Intrusion Detection Based On Neural Network Model

Nowadays,people's lives are surrounded by the Internet,and network equipment needs to operate uninterruptedly to meet the needs of end-users.A large amount of data is exchanged with network equipment through public and private networks at all times,but the Internet and network equipment are open connected access brings network security threats in most daily network activities.The network intrusion detection system is a reliable and efficient technology to ensure network security.However,the current network intrusion detection system still has the problems of low detection accuracy and high false alarm rate when facing imbalanced samples.The network intrusion detection based on neural network model studied in this paper has important theoretical significance and application value.The main research contents of this paper are as follows:(1)Summarize and analyze the current network security and information security issues,as well as the current research status of intrusion detection technology at home and abroad.The classification of intrusion detection technology and the practical application of network intrusion detection technology are introduced,the structure of neural network and the principle of learning features are analyzed.(2)Aiming at the problem that the neural network model is biased to the majority of samples when learning the characteristics of imbalanced samples,which leads to the overfitting of the model,a network intrusion detection method based on 1DCNN-GRU is proposed.The 1DCNN-GRU hybrid model is a deep network model composed of One-Dimensional Convolutional Neural Networks(1DCNN)and Gate Recurrent Unit(GRU).The convolutional layer is convolved by dilated convolution to extract data features,the GRU layer uses reset gates and update gates to further filter the feature data,remove redundant information,and retain feature information with category differentiation.Adding a Gaussian Noise layer during model training increases sample diversity,improves the learning ability of the model,and improves the expression ability of the model by adjusting the optimizer and learning rate of the model.Simulation experiments prove that the detection performance of the 1DCNN-GRU model has been improved.(3)In order to solve the problem that the neural network model cannot be fully learned due to the imbalance of the sample,a network intrusion detection method based on SENNBys-GRU is proposed.Use SMOTEENN algorithm(SENN)to combine random oversampling minority class and random under-sampling majority class to synthesize new sample data.Use Bayesian optimization algorithm(Bys)to find a set of optimal hyperparameters for GRU network model,and finally use the GRU network model for detection and classification.Simulation experiments prove that the SENN-Bys-GRU model can effectively improve the accuracy of detection and reduce the false alarm rate of detection.The innovations of the research work in this paper are mainly reflected in the following aspects:(1)A new intrusion detection method 1DCNN-GRU is proposed.The GRU network layer is placed between the 1DCNN layer and the fully connected layer,making full use of the advantages of the two neural networks to improve the learning ability of the model.During model training,Gaussian Noise is added to the input samples to increase the diversity of the samples,thereby improving the model's ability to learn mapping rules from the input space and enhancing the robustness of the network.(2)A network intrusion detection method based on SENN-Bys-GRU is proposed.The SMOTEENN algorithm synthesizes minority samples and eliminates noise samples in the majority,which solves the problem of insufficient samples and noisy samples.It takes a lot of time and effort to search and try the optimal hyperparameters with artificial full probability,While the Bayesian optimization algorithm can obtain the optimal hyperparameter combination by only a small amount of evaluation of the objective function through the probability model and the acquisition function.