Research And Implementation Of Intrusion Detection Method Based On Deep Learning

As the high-speed development of Internet technology,network security has become more important.Recently,there have been frequent cyber attacks,which have had a great impact on daily life.As the key technology of network security protection,intrusion detection technology is very important.However,there are still two general problems in its current research.The first problem is that many studies ignore the false alarm rate(FAR)and cause too much warning information in the actual use of the network intrusion detection system.The second problem is that there are very few studies on the multi-classification of invading virus types.The multi classification intrusion detection model can not only provide alarm information,but also provide attack type information,so that the system can take countermeasures directly.This thesis introduces the attention mechanism into the field of network intrusion detection and conducts detailed research.First of all,this thesis proposes a LSTM intrusion detection model combined with self-attention mechanism.Experiments on the CICIDS2017 data set show that the accuracy of the model's fifteen traffic classifications is as high as 99.591%.While the FAR is only 1.127%.It realizes that while improving the detection accuracy of the LSTM intrusion detection model,and it also reduces the false alarm rate.Then,further using multiple attentions to splice together,this thesis proposes an LSTM intrusion detection model based on multi-head attention mechanism,which realizes the performance enhancement of single-head attention mechanism by focusing on different parts of multiple attentions.In the simulation experiment,the accuracy of the fifteen traffic classifications of this model is 99.629%,and the FAR is only 0.991%.Compared with the self-attention mechanism,it has a higher accuracy rate and a lower false alarm rate.Finally,this thesis further explores the performance comparison of different models combined with the attention mechanism,and finds that the model combining the LSTM and the attention mechanism proposed in this thesis has higher detection capabilities.Then,the intrusion detection model proposed in this thesis is compared with the commonly used deep learning models such as convolutional neural network in intrusion detection,and it is found that the model proposed in this thesis has stronger detection accuracy and lower FAR.