Traffic Anomaly Detection for Data Communication Networks

Abstract

The detection efficiency of the traditional data communication network traffic anomaly detection algorithm is low. And it is impossible to guarantee the accuracy of traffic detection in actual applications. The detection algorithm involves too many dimensions, and it is difficult to explore the optimal solution even if it takes a lot of time. In view of the above problems, this paper proposes an improved network traffic anomaly detection algorithm. The algorithm inherits the algorithm idea of combining the weak classifiers in the classical GBDT (Gradient Boosting Decision Tree) into the final strong classifier. The algorithm equilibrium weights are assigned to the weak classifiers in the iteration to balance the contribution of the weak classifier to the final classification model. The algorithm combines Bayesian optimization algorithm to achieve the purpose of automatically exploring the optimal super-parameter combination. The simulation results show that the proposed algorithm has an effective improvement in detection efficiency compared with the traditional traffic detection algorithm.