| With the development of information technology,more and more attention has been paid to Cyberspace Security.Intrusion detection technology has become the backbone of defense against Cyberspace Security Intrusion.However,the performance of existing intrusion detection system is generally problematic,and depends on manual maintenance and update.Machine learning combines probability statistics and mathematical analysis,it has attracted a lot of attention in recent years.Combining machine learning algorithm with intrusion detection technology to solve the existing problems is a very promising research topic.Machine learning methods are very dependent on the training set.However,it is difficult to collect the real attacks in the normal traffic in the network environment.Firstly,malicious attacks are submerged in a large number of normal traffic,and only a small part of valuable attack data can be separated,which causes the low data utilization rate.Secondly,there are many kinds of malicious attack samples,and requires manual classification and selection.Aiming at the problem of low data utilization,this paper proposed a method of generating malicious attack samples used machine learning generation model.In this paper,Generation Adversarial Network was used as the basic framework of the generation model,and the structure of Long Short Term Memory network was applied to the network structure.In the gradient back propagation,the strategy gradient method was used to make the training results more stable.At the same time,the experiment was designed to prove that the generated samples of the model have the real sample characteristics.The experimental results have shown that the generation model works well.This paper designed an attack sample collection system based on honeypot technology and solved the problem of further classification of samples.It completed the attack sample collection by targeted deployment of specific attack sample. |
PDF Full Text Request