Um modelo de reputação com classificação via agrupamento para detecção de nós egoístas em redes oportunistas

Abstract

The advance of networking infrastructure technologies has spawned a new range of multi-hop applications, such as sensor networks and wireless mesh networks. The concept of delay and disruption-tolerant networks has emerged as a solution to enable communication in scenarios in which communication is intermittent. However, a basic premise for the good performance of these networks is the collaboration of the nodes during the communication. Such collaboration is a critical issue to the data flow and overall performance. However, due to factors resources constraints (e.g., buffer and energy), the nodes can act selfishly, and stop collaborating with the data flow in the network. The existence of a distributed mechanism to measure the collaboration degree of the network nodes is, therefore, vital to improve the delivery of messages. This work proposes a model of egoism detection using reputation mechanisms, which can qualify other members of the network according to their participation degree in the data flow to identify both selfish and cooperative nodes. The proposed model makes use of a numerical-based ranking method through executed during the contacts. Once qualified, a grouping technique classifies the network nodes, which differs from other works found in the literature. The experimental results, got in the simulation environment The ONE, show that the model proposed and implemented here is promising and the clustering technique can be applied without loss of reliability, and being very precise when the rate of selfish nodes in the network increases.