Research On The Application Of Deep Learning In Intrusion Detection

Computer networks are convenient for people,but due to the lack of security features and the openness and interconnectivity of the network itself,the network is vulnerable to hackers,computer viruses,malware and other misdemeanors.As an active defense technology,intrusion detection plays an important role in the field of network security.Traditional intrusion detection technology faces the problems of low accuracy,low detection efficiency and high false alarm rate.How to improve the accuracy and efficiency of intrusion detection and reduce the false alarm rate of intrusion detection is still an important issue for security personnel.In this paper,an intrusion detection model based on deep learning and probabilistic neural network is proposed.Combined with the optimization concept of particle swarm optimization algorithm,the particle swarm algorithm is used to determine the structure of the deep automatic encoder and the spread constant of the probabilistic neural network to improve the efficiency and accuracy of intrusion detection.Main research contents include:(1)Intrusion detection data analysis and prepossessing.It analyzes the characteristics of intrusion detection data,compares the difference and relation between intrusion detection data set KDD99 and NSL-KDD,then intrusion detection data is prepossessed,including data cleaning,coding,and integration of extraction and normalization.(2)Probabilistic neural network intrusion detection model and optimization.Focus on the spread constant of probabilistic neural network,the use of particle swarm optimization algorithm to find the optimal value,in order to improve the classification of probabilistic neural network.The model was constructed by Matlab,and to verify the effectiveness of PSO,the performance of the model was tested by using the data set.(3)Intrusion detection model based on deep learning and probabilistic neural network: Combined with the deep learning theory,the use of deep automatic encoder to extract feature of experimental data.Through the comparison of many experiments,the number of DAE iterations and the number of hidden layers were determined.To solve the problem of determining the number of hidden layer nodes in DAE,a particle swarm optimization algorithm is used to find the optimal number of hidden layer nodes.Then,an intrusion detection model based on deep learning and probabilistic neural network is constructed.Through training and testing of the model,it shows that this method effectively improves the speed of intrusion detection.(4)Build and test the prototype system: Build a prototype system to test the feasibility of the intrusion detection model based on deep learning and probabilistic neural networks in a real network environment.By simulating cyber attacks,it is shown that the model is feasible.