Intrusion Detection Analysis And Application Based On XGBoost Algorithm

In recent years,the development of network technology has made the explosive growth of the Internet.With more and more network services,network security issues are getting more and more attention.From the 80 s of last century,In response to cyberthreats,cybersecurity has been put into research.Many classic network security mechanisms such as digital signatures,network encryption layer,access control list technology and firewall technology,etc.,in the past the network with less change,good network security can be achieved.However,in today's network environment,frequent changes in network status have made the abovementioned static protection measures worse and worse.Therefore,the introduction of intrusion detection system.Intrusion detection system and the traditional network security technology is different from the performance point of view,intrusion detection system can reduce the overall false alarm rate.For unknown attacks,Intrusion detection system to respond quickly,take the initiative to learn the new threat characteristics,to ensure that the decline in false negatives.Intrusion detection system is the most important part of the detection and analysis of the method used.In recent years,machine learning technology has made considerable progress,so this article from the introduction of decision tree algorithm,the principle of such machine learn-ing algorithm expand narrative.At the same time,the advantages of the decision tree algorithm in intrusion detection system in machine learning algorithm are described and the performance of the algorithm is simply evaluated.After the decision tree algorithm based on the introduction of the latest XGBoost algorithm.Compared with the basic decision tree algorithm,XGBoost algorithm introduces L1 regularization and L2 regularization in regularization optimization,and adds regular items in the cost function to control the complexity of the entire model and prevent over-fitting.From the perspective of cost function,the gradient-ascending tree algorithm uses only the first derivative,and XGBoost algorithm Taylor expansion,which takes the second derivative term.XGBoost algorithm as a lifting tree algorithm One of the lower right of each iteration of the resulting decision tree heavy,this method can effectively reduce the impact of each tree was the overall model,decision tree generated after training in greater learning space.At the same time,XGBoost algorithm draws on the column sampling tech-nique in random forest,which can reduce the amount of computation and inhibit the over-fitting phenomenon.As an optimized tree algorithm,the XGBoost algorithm not only greatly improves the accuracy,but more importantly,it optimizes the project and adopts the parallel computing method to increase the speed.When using the XGBoost model for training,the model provides parallel training on feature granularity.Firstly,the feature data is sorted first,and then distributed to different nodes through distributed principle,so that can greatly improve the overall efficiency.This article concludes by providing an optimal XGBoost model for the KDD Cup 99 in a network dataset and evaluating its performance.