Simulation Research on Camouflaged Intrusion Detection Model Based on Improved RF Algorithm

It is of great significance to detect and stop camouflage intrusion quickly and effectively for maintaining the security of network resources. As an important technical means to maintain network security, camouflage intrusion security detection has attracted more and more attention. RF (Random forest) is a method that combines a series of decision subtrees, and then determines the final output result of samples according to the voting results of subtrees. In this paper, the simulation research of camouflage ID (intrusion detection) model based on improved RF algorithm is carried out. In this paper, an improved RF algorithm is proposed. Firstly, the mixed sampling method is adopted before RF is divided into subsets to reduce the imbalance of each subset. Then, the subtree with strong independence is selected from many decision trees to form the final RF. In the process, K-means algorithm is used to cluster the decision trees and extract the cluster center. The research results show that with the increase of information collection duration, the detection algorithm will show better performance. When the information collection time is 50 minutes, the system shows the best performance (ACC: 83.106%, FP: 12.615%). The improved RF algorithm can better eliminate the influence of redundant features on the model, and has better processing effect on unbalanced data.