Intrusion Detection Classification Based on Random Forest

Abstract

In recent years, network security has become increasingly severe and network intrusions are frequent. On the one hand, it is difficult to capture intrusions because of the increasingly large volume of computer networks and complex network topology, and on the other hand, it is more difficult to capture intrusions because the characteristics of intrusions are becoming more diverse and complex. The purpose of this paper is to study intrusion detection classification based on random forest. After constructing a high-precision decision tree, the double failure metric is used as the distance between two decision trees, and the k-means++ algorithm is used to select high-precision decision trees with a certain degree of independence to form the final random forest. The improved random forest algorithm for setting up clustering and the random forest algorithm were experimented as algorithm comparisons. The KDD-NSL dataset was selected and the experimental results were compared to demonstrate the effectiveness of the random forest algorithm based on gradient boosting. The experimental results show that the improved random forest model has a significant reduction in decision tree size and a significant increase in diversity due to the k-means++ algorithm for clustering and extracting cluster centers.