Grant: Discovery Grants Program - Individual

One for all, all for one. Distributed systems are generally composed of independent participants who have to collaborate with each other for their own benefit. Without a strong cooperation among the participants, these systems would be ineffective. Unfortunately, participants cooperate legitimately, act maliciously or behave selfishly. For example, in mobile ad hoc networks, the participants may have to forward traffic flows from others. Thus, selfish participants can be seen as passive participants who retrieve some network services from others without providing back any such service. These participants can have major impacts on the performance of such a network. Ultimately, legitimate participants may not be able to communicate properly.Without any incentive mechanism, selfish participants would behave for their own benefit without any regard for others. Traditionally, game theory has been used to model incentive mechanisms to reward legitimate participants with a faked currency. Repeated non-cooperative game approaches can also be used to force selfish participants to collaborate under the threat of potential retaliation. Unfortunately, traditional approaches assume that cooperative participants can always collaborate truthfully. They cannot cope easily with stochastic participants who are unfortunately unable to collaborate momentarily. Similarly, traditional approaches assume that cooperative and uncooperative participants can always be distinguished without any error. If a participant misidentifies another participant, this latter participant may be facing unjustifiable retaliations. This research program would investigate how to adapt the classical approaches of game theory (i) to allow the presence of stochastic cooperative participants and (ii) to cope with misidentification of the cooperative/uncooperative participants.