Research On Intrusion Prevention Technology Based On Three-Way Decisions And Deep Learning

Society is developing and technology is advancing.The rapid development of the Internet has brought the Internet into a rapidly changing era.The rapid development of network technology has brought people a convenient life experience,but at the same time it also brings a huge threat.Some issues are related to network security and have become key issues that need to be resolved in this era.Firewalls and intrusion detection technologies are common technical equipment to protect network security,but the network environment is more complex than before,and intrusion methods are also being upgraded.Therefore,the effects achieved by these commonly used technologies have been difficult to satisfy users.Intrusion prevention technology absorbs the desirability of firewall and intrusion detection,and has gradually become a new generation of network security technology.With the deepening of research on artificial intelligence technology,algorithms such as machine learning are increasingly used in intrusion prevention systems to detect network behaviors.However,the existing intrusion detection algorithms based on machine learning are mostly based on two-way decisions,that is,a network behavior should be immediately recognized as an abnormal behavior or a normal behavior without any delay.Three-way decisions theory firstly is introduced,and due to the advantages of deep learning in feature extraction and data generation,deep learning methods and three-way decisions are combined and used in intrusion detection.An intrusion prevention model based on three-way decisions and deep learning models is constructed.Firstly,in view of the shortcomings of the current existing machine learning algorithms in the field of intrusion detection,a multi-granularity intrusion detection algorithm DAE-3WD based on denoising autoencoder and three-way decisions is proposed.The multi-granularity feature set is constructed through the non-linear feature extraction method of autoencoder,and original intrusion detection problem is decomposed into multiple sub-problems through three-way decisions theory,and each feature set corresponds to the information needed by a classifier model to solve a sub-problem.Experiments proved the effectiveness of DAE-3WD model.Furthermore,in order to distinguish different intrusion behaviors,an intrusion detection algorithm based on variational autoencoder and three-way decisions is proposed.Variational autoencoder,a deep generative model,is used to generate new behavior data based on existing network behavior data,and the generated samples would be used as a training set to train the classifier model.The variational autoencoder model is combined with the three-way decisions theory to make it generate new samples based on the existing data at an appropriate time.The experiment proves the effectiveness of the VAE-3WD model.Finally,the deployment of the intrusion prevention system was designed,and simulation experiments are carried out by related test tools.The experimental results proved the feasibility of the model based on three-way decisions and deep learning in intrusion prevention.