Study of Recognition Algorithm for Core Node in KAD Network Based on BP Model

Abstract

Abstract-In view of the core node recognition in the KAD (Kademlia), a model based on BP is presented to determine whether a node is core node in real time. According to the result of the measurement for KAD, some attribute characteristics of core nodes in the network have been extracted and normalized to obtain an attribute set with higher separable degree. An algorithm in MatLab is designed to train the BP network until the results are limited in a predetermined error range. In addition, the trained BP model is adopted to determine the tested nodes, and the results of the experiment show that the method can judge degrees of importance of nodes in real time, and the recognition accuracy rate is up to about 70%.