BRACONNOT-VELLOSO Pedro

doctorant à Sorbonne Université
Équipe : Phare
https://lip6.fr/Pedro.Velloso

Direction de recherche : Guy PUJOLLE

Co-encadrement : DUARTE Otto

Un modèle de confiance pour les réseaux ad hoc

This thesis addresses the problem of trust evaluation and management in ad hoc networks, in which nodes accumulate the role of router, server, and client compelling them to cooperate for the correct operation of the network. Several new protocols and applications have been proposed and developed because traditional solutions are not adequate for ad hoc networks. Most of the proposed works, however, considers the perfect cooperation among all nodes assuming that they all behave according to the application and protocol specifications. Nevertheless, this assumption may be false, due to resource restrictions or malicious behavior. Eventually, this unexpected behavior can degrade the network performance, increase the resource consumption, and augment vulnerability to attacks. Therefore, a mechanism that allows a node to infer the trustworthiness of other nodes is necessary. We propose a trust model based on the concept of human trust. The model builds a trust relationship among the nodes of an ad hoc network based on previous experience and recommendations. We present the Recommendation Exchange Protocol (REP) which allows nodes to send and receive recommendations of its neighbors. The goal is to make nodes capable of gathering information to reason, learn, and make their own decisions. We focus on providing nodes with a trust level for each direct neighbor, that is, a neighbor within the radio range. Different from most related works, our work scales well for large networks by restricting nodes to keep and exchange trust information solely with direct neighbors. This characteristic decreases the number of messages, and consequently, the energy consumption. In addition, it helps to mitigate the effect of colluding attacks of liars in the network. We also introduce the concept of relationship maturity which allows node to improve the efficiency of the proposed model in mobile scenarios. We show the correctness of our model in a single hop network through simulations in a simulator developed specifically for our model. Then, we extended the analysis to mobile multi-hop networks, showing the benefits from using the maturity relationship concept. At last, we evaluate the impact of malicious nodes that send false recommendations to degrade the efficiency of the trust model. The results show that our model tolerates up to 40% of malicious nodes.

Soutenance : 18/07/2008

Membres du jury :

Maryline Laurent-MAKNAVICIUS [Rapporteur]
Marcelo G. RUBINSTEIN [Rapporteur]
Aline C. VIANA
Laurent REYNAUD
Guy PUJOLLE
Otto Carlos M. B. DUARTE

Date de départ : 30/06/2009

Publications 2006-2023