期刊名称：International Journal of Computer Trends and Technology
电子版ISSN：2231-2803
出版年度：2016
卷号：40
期号：1
页码：5-9
DOI：10.14445/22312803/IJCTT-V40P102
出版社：Seventh Sense Research Group
摘要：MANET consists of self organized mobile nodes. For effective maintenance of network, these mobile nodes should collaborate together. But, some nodes in the network may fail to cooperate and show selfish behavior. The presence of selfish nodes in the network reduces the performance of packet transmission. Watchdog systems are used in the detection of selfish nodes. Hence, collaborative contactbased watchdog (CoCoWa) approach was proposed by which selfish nodes, new contacts and malicious nodes were detected by using local watchdogs. This approach enhanced the effectiveness of detecting selfish nodes. In our proposed system, malicious nodes are detected by periodically monitoring the behavioural changes of each node. The observed changes are used to construct the behaviour matrix. KullbackLeibler divergence (KL divergence) is then used to calculate the divergence rate of nodes from behaviour matrix and identifies the malicious nodes. But, when the argument values used in KL divergence are reversed, then it show different results because is a nonsymmetric divergence. This results leads to the accuracy problem. Hence, our proposed work uses different forms of mixture of directed and alternate Kullback–Leibler divergences for symmetrizing the Kullback–Leibler Divergence. Based on the obtained symmetric measures, Akaike information criterion (AIC) is then used for improving accuracy results by selecting the malicious nodes effectively.
关键词：Watchdog system; malicious nodes; Kullback-Leibler divergence and Akaike informationcriterion
