Research On Intrusion Detection Method Based On Deep Learning

The Internet has become an indispensable part of people’s daily life.When people enjoy the convenience brought by the Internet,they are also faced with network security problems.As a network security technology to identify and deal with network intrusion,intrusion detection plays an important role in protecting network security.Traditional intrusion detection methods have the problems of low detection rate,inability to identify new attacks,and the imbalance of intrusion detection data,which makes the classification results biased and affects the overall performance of the intrusion detection model.In view of the above problems,this thesis studies the intrusion detection method based on deep learning.The main research contents and problems to be solved are as follows:(1)Aiming at the problems of low detection rate and high false alarm rate in existing intrusion detection methods,this thesis proposes an intrusion detection method based on VAE-CNN-Bi LSTM.This method first uses the VAE to learn the internal characteristic distribution of network traffic data,thus generating data similar to the original network traffic data,enriching the temporal and spatial characteristics of network traffic data,Then,CNN and Bi LSTM are used to extract the spatial and temporal characteristics of network traffic data.The simulation experiment on NSL-KDD data set shows that the detection accuracy of VAE-CNN-Bi LSTM reaches 99.29%,and the detection accuracy reaches 99.64%.(2)In order to solve the problem that a few classes of samples are difficult to identify due to unbalanced data in intrusion detection,this thesis proposes a mixed sampling method based on Borderline-SMOTE and NCL.This method first uses Borderline-SMOTE algorithm for oversampling,and then uses NCL algorithm for undersampling to improve the imbalance of data.The simulation experiment on NSL-KDD data set shows that the mixed sampling method effectively improves the overall detection effect and improves the detection rate of a few samples.The recall rate of R2 L in NSL-KDD data set reaches 94.63%,and the F1-score of U2 R in NSL-KDD data set reaches 95.13%.To sum up,the research work in this thesis has solved some problems existing in current intrusion detection,not only improving the overall performance of intrusion detection,but also improving the detection effect of unbalanced data.