Research On Anomaly Traffic Detection Method Based On Improved AdaBoost Algorithm

With today's society entering the Internet era,network applications have gradually integrated into our daily lives.due to the explosive development of the network,there are subsequent security issues,many networks exist or are latent with a variety of network attacks or network security threats.Therefore,as an important part of resisting network attack,abnormal traffic detection provides an initial shield for network security.Providing accurate abnormal traffic detection is necessary to ensure network security.This article studies the abnormal flow detection technology by reading a large number of literatures,analyzes the characteristics of statistical abnormal flow from various angles,as well as the related work on abnormal flow.The machine learning algorithm was selected,and the number of iterations of the original Adaboost algorithm was controlled by combining the momentum method with the Adaboost algorithm and setting the threshold to eliminate redundancy.The improved-Adaboost algorithm based on the Adaboost algorithm was proposed and an abnormal flow detection model was designed and implemented for experimental verification.The whole experiment is divided into three parts: first,after preprocessing the KDD CUP99 data set,the first part of the improved-Adaboost_algorithm and Adaboost algorithm,Randomforest algorithm,GradientBoosting algorithm,DecisionTree algorithm,Extratree algorithm in accuracy,recall rate,F1 to test the performance of the algorithm,to evaluate the better algorithm.In the second part,this paper visualizes the classification process of each algorithm,and describes the role of each algorithm in the model through the form of tree graph,so as to know the advantages and disadvantages of the algorithm from multiple perspectives.In this paper,we try to visualize the range of the extreme value of each classification algorithm,and show the range of the extreme value of each algorithm in the form of scatter graph.Finally,experimental data can improve the improved-Adaboost algorithm in abnormal flow detection.