The main principle behind the working of delay tolerant networks (DTN) is the mobility of the nodes along with their contact sequences for exchanging data. Nodes which are a part of the DTN network can behave selfishly due to network reservation policy, especially when constrained to energy or storage space. Several forwarding protocols exist for spreading data but our focus is on the performance of popular data relaying protocols namely epidemic routing and two hop routing protocol in a situation where nodes exhibit various degrees of selfishness. Results of an analytical model show the performance advantage of epidemic routing over two hop routing decreases as the number of selfish nodes and intensity of the selfishness increases either deterministically or probabilistically. We practically asses the vulnerability of the above mentioned protocols using ONE simulator. We find that our result coincides with analytical results with some variations in the graph.

In Mobile Ad Hoc Networks (MANETs), nodes depend on each other for routing and forwarding packets. However, to save power and other resources, nodes belonging to independent authorities may behave selfishly, and may not be willing to help other nodes. Such selfish behavior poses a real threat to the proper functioning of MANETs. One way to foster node cooperation is to introduce punishment for selfish nodes. Based on neighbor-monitoring techniques, a fully distributed solution to detect, punish, and re-admit selfish nodes, is proposed here. This solution provides nodes the same opportunity to serve/and be served by others. A light-weight solution regarding battery status is also proposed here. This solution requires neighbor monitoring only when necessary, thereby saving nodes battery power. Another effective way to solve the selfish-node problem is to reward nodes for their service according to their cost. To force nodes to show their true cost, truthful protocols are needed. A low overhead truthful routing protocol to find optimal routes is proposed in this thesis. The most prominent feature of this protocol is the reduction of overhead from existing solutions O(n3) to O(n2). A light-weight scalable truthful routing protocol (LSTOP) is further proposed, which finds near-least-cost paths in dense networks. LSTOP reduces overhead to O(n) on average, and O(n2) in worst case scenarios. Multiple path routing protocols are an effective alternative to single path routing protocols. A generic mechanism that can turn any table-driven multipath routing protocol into a truthful one, is outlined here. A truthful multipath routing protocol (TMRP), based on well-known AOMDV protocol, is presented as an example. TMRP incurs an only 2n message overhead for a route discovery, and can also achieve load balancing without compromising truthfulness. To cope with the selfish-node problem in the area of position-based routing, a truthful geographic forwarding (TGF) algorithm is presented. TGF utilizes three auction-based forwarding schemes to stimulate node cooperation. The truthfulness of these schemes is proven, and their performance is evaluated through statistical analysis and simulation studies.

Les réseaux ad hoc véhiculaires (VANETs) permettent le partage de différents types de données entre les véhicules, de manière collaborative. Dans cette thèse, nous nous sommes tout particulièrement intéressés aux applications de sûreté et de sécurité routière, dédiées à l'échange des informations sur l'état de l'environnement routier. Les contraintes de ces applications en termes de qualité de services sont des plus rigoureuses, car l'acheminent de leurs données doit être exhaustif et ne souffrir d'aucun retard pour assurer une information utile et en temps opportun au profit de tous les usagers concernés. Cet acheminement doit faire face aux difficultés induites par la dispersion et la forte mobilité des véhicules, l'absence ou l'insuffisance d'infrastructure, la densité variable du réseau, la surcharge en informations à envoyer et l'étendue des zones géographiques à couvrir. En effet, la problématique de diffusion des données dans les VANETs s'avère non-triviale et de nombreux verrous scientifiques doivent être levés pour permettre un support efficace, collaboratif et fiable pour les applications de sûreté et de sécurité routière.Plus précisément, nous aborderons la problématique de la dissémination collaborative en se posant trois questions : “comment disséminer les données ? À quel moment le faire ? Mais aussi quoi disséminer et comment inciter à le faire ? ” Nous avons apporté des réponses à travers les trois contributions de cette thèse. La première consiste à proposer une stratégie de dissémination efficace, qui soit adaptée à l'importance de l'information échangée et à sa durée de vie, permettant ainsi d'éviter un processus de diffusion intensif. Celui-ci est inapproprié dans ce cas de figure, car il génère de la congestion et beaucoup de redondance. Une étude de performances par simulation est réalisée, laquelle montre une diminution de 90% du taux de messages redondants par rapport au cas de la diffusion par inondation. Afin d'améliorer plus encore les performances du processus de diffusion des messages de sûreté, nous proposons, dans un second temps, un ordonnanceur pour l'accès au canal de communication qui a pour objectif de réduire le nombre de collisions dues aux synchronisations afférentes à l'utilisation du multi-canal dans le standard IEEE 802.11p/1609.4 et donc élever le taux de réception des données. Nous basons notre proposition sur la théorie de l'arrêt optimal, qui décide du moment opportun pour l'envoi d'une information en conciliant occupation du canal, efficacité de l'envoi et délai d'ajournement toléré par une information. Dans notre cas, la théorie de l'arrêt optimal est formulée par un processus de décision Markovien (MDP). Nous montrons ainsi par simulation une amélioration substantielle du taux de réception (de 25%) et une diminution importante des pertes (de 47%).Après s'être intéressé à l'aspect quantitatif des performances du réseau, nous nous intéresserons ensuite à l'amélioration de la fiabilité du processus de diffusion. Cette fiabilité est obtenue grâce à l'incitation des véhicules à la coopération et à l'exclusion des véhicules malicieux de celui-ci. Ceci est réalisé au travers de la proposition d'un modèle de confiance, inspiré des jeux de signaux. Le modèle crée une situation d'équilibre, tel que les différentes parties le composant ne soient pas tentées de le contourner, ainsi découle une auto-sélection des véhicules, laquelle est rapide et peu coûteuse. À notre connaissance, notre modèle est le seul à s'attaquer aux effets néfastes des deux types de véhicules, malicieux et égoïstes, en même temps. Comme précédemment, nous évaluons les performances de notre solution au travers d'une modélisation par une chaîne de Markov et divers jeux de simulation. Ceci a permis de montrer que 100% des véhicules malicieux sont exclus, avec le maintien d'un taux de coopération élevé dans le réseau, soit une amélioration de 42% / Vehicular Ad Hoc Networks (VANETs) allow sharing different kinds of data between vehicles in a collaborative way. In this thesis, we are particularly interested in road safety applications, designed for the exchange of information on road traffic and conditions. This kind of applications have strict Quality of Service (QoS) requirements, as data must be routed thoroughly and without any delays so for assuring the timely delivery of useful information to the drivers. In this context, data routing must face several issues raised by the high mobility and dispersion of vehicles, inadequate or completely lacking infrastructure, a variable network density, network saturation due to the large of information to deliver, and the size of the geographical areas to cover. Indeed, the problem of data dissemination in VANETs is non-trivial, and several research challenges must be solved in order to provide an efficient, collaborative, and reliable support for road safety applications. Specifically, we will address the problem of collaborative data dissemination through the following three questions: “How to perform data dissemination?”, “When should we do it?”, and “What must be disseminated?” We have provided answers to these questions through the three contributions of this thesis. Our first contribution is an efficient dissemination strategy, specifically tailored to the importance of the exchanged information as well as its lifespan, which is able to avoid the intensive dissemination process that generates network congestion and data redundancy. We confirm our statements and validate the performance of our solution by modeling it using a discrete-time Markov chain, which demonstrates the number of necessary retransmissions for all concerned vehicles to receive information. Moreover, we performed extensive simulations that show a reduction of up to 90% of redundant messages with respect to message flooding dissemination strategies. Next, in order to further improve the road safety message dissemination process, we propose a communications channel access scheduler, which aims at reducing the number of collisions caused by IEEE 802.11p/1609.4 multi-channel synchronizations, and thus improving the data reception rate. We base our solution on the optimal stopping theory, which chooses the right moment to send information by balancing the channel occupancy rate, the data delivery efficiency, and the maximum deferment delay tolerated by the information. To this end, we formulate the optimal stopping theory through a Markov decision process (MDP). We show through simulation-based evaluations an improvement of the reception rate of up to 25% and a reduction of up to 47% of message losses. Finally, after being interested in the quantitative aspect of network performance, we centered our efforts on improving the reliability of the dissemination process, which is obtained by motivating vehicles to cooperate and evicting malicious vehicles from the process. To this end, we propose a trust model inspired on signaling games, which are a type of dynamic Bayesian games. Through the use of our model, equilibrium is achieved, thus resulting in a fast and low-cost vehicle self-selection process. We define the parameters of our trust model through a discrete-time Markov chain model. To the best of our knowledge, our solution is the only existing solution that tackles the negative effects introduced by the presence of both malicious and selfish vehicles in a VANET. We evaluated the performance of our solution by modeling it using a Markov chain, and a set of simulations. Our results show that up to 100% of malicious vehicles are evicted while keeping a high cooperation rate, thus achieving an improvement of 42% when compared to other similar solutions