Research On Multi-node Ensemble Algorithm And Application On IDS

BP neural network can be used to make intrusion detection, but the recognition rate is low. AdaBoost ensemble algorithm can improve the recognition rate by ensembling BP neural networks which have differences.But with the rapid development of the network, in the face of massive data, the traditional serial computing can not meet the requirements of fast training. Therefore, the demand for multi-node computing in the network environment is increasing. This paper mainly deals with the research on a multi-node ensemble algorithm and the application on IDS.Firstly, through having some knowledge of BP neural network, a BP neural network is used as the base classifier of AdaBoost algorithm. Then AdaBoost algorithm is introduced in detail, and a parallel AdaBoost algorithm is improved. The distribution of the samples, which is adjusted by the classifier with the best performance in the original algorithm, is adjusted by the train error of the classifiers obtained each round in the improved algorithm. Then a multi-node AdaBoost algorithm is proposed. This algorithm is applied on the computer cluster of the multi-node environment, using FCM to implement a algorithm to adjust the sample distribution, so that multiple classifiers can be trained simultaneously on multiple nodes. Finally, the improved multi-node parallel AdaBoost algorithm and the multi-node AdaBoost algorithm based on FCM algorithm are applied to the intrusion detection. In this system, we make the feature extraction to the intrusion detection data first. The pretreated ID data will be the data sets of the experiment. Then we compare the improved AdaBoost algorithm and the traditional AdaBoost algorithm by using the data sets, experiment shows that the improved algorithm has better performance.