Research On Intrusion Detection System Model Based On Deep Neural Network

At present,network technology has been developed rapidly.Network applications appear in every corner of society,but at the same time,network security problems are also becoming more and more important.The emergence of new types of network attacks has made them very threatening.In response to this problem,the Intrusion Detection System(IDS)is currently used for defend against network attack.IDS has become the hottest research project because of its reliability,expandability and self-learning.However,at present,IDS generally have problems such as low recognition accuracy,weak generalization ability and so on.Therefore,the paper introduces the deep neural network in the artificial intelligence algorithm.The paper also proposes an improved multi-scale convolutional neural network to completely extract the spatial characteristics of the network traffic,and proposes an intrusion detection system based on deep neural network to solve the above practical problems.The main work and innovation are as follows:1.The paper introduces the knowledge of artificial intelligence algorithm and intrusion detection system,introduces the research status at home and abroad.The performance characteristics and defects of classical algorithms are discussed in the paper.The paper also introduces deep neural network into intrusion detection system and analyzes how this new algorithm successfully solves the problems of low detection rate and weak generalization ability of the classical model.2.In the paper,a multi-scale convolutional neural network algorithm is proposed(MSCNN).The paper explores how to apply neural network frameworks to the field of network intrusion detection.The paper also designs an experiment and fully analyzes the advantages of MSCNN algorithm in CNN improved algorithm.MSCNN is more suitable to establish the whole system,which makes up for the defects of low detection accuracy,incomplete feature extraction and mismatch of input and output when the CNN network is used in the intrusion detection field.3.In the paper,we propose an intrusion detection system based on multi-scale convolutional neural network and long-term memory network fusion detection algorithm named MCL-IDS.The model first uses a convolutional neural network with multiple convolution kernels of different scales to process the spatial features of the data,and then combines a long-term and short-term memory network to process the temporal features,to complete the classification of network traffic from the point of view of optimal data features.The paper introduces the function of each stage of the model,and designs the simulation with the experiment.The simulation results show that the proposed MCL-IDS model effectively improves the accuracy of intrusion detection,and reduces thefalse positive rate and the false negative rate.4.The paper builds the experimental platform,implements the MCL-IDS model,and selects a new intrusion detection data set UNSW-NB15 as the basis to build the large-scale network flow characteristic database needed by the training algorithm.In the paper,we load the MCL-IDS model on the platform,and design a control experiment against rare intrusion attacks in the actual network environment,which verifies that the generalization ability of the MCL-IDS model is better than the classical model.