A Game Theoretic Trust Model for On-Line Distributed Evolution of Cooperation in MANETs Marcela Mejia a,* , N´ estor Pe˜ na a , Jose L. Mu ˜ noz b , Oscar Esparza b , Marco A. Alzate c a Universidad de los Andes, Colombia b Universitat Polit` ecnica de Catalunya, Spain c Universidad Distrital, Colombia Abstract Cooperation among nodes is fundamental for the operation of mobile ad hoc networks (MANETs). In such networks, there could be selfish nodes that use resources from other nodes to send their packets but that do not offer their resources to forward packets for other nodes. Thus, a coop- eration enforcement mechanism is necessary. Trust models have been proposed as mechanisms to incentive cooperation in MANETs and some of them are based on game theory concepts. Among game theoretic trust models, those that make nodes’ strategies evolve genetically have shown promising results for cooperation improvement. However, current approaches propose a highly centralized genetic evolution which render them unfeasible for practical purposes in MANETs. In this article, we propose a trust model based on a non-cooperative game that uses a bacterial-like algorithm to let the nodes quickly learn the appropriate cooperation behavior. Our model is completely distributed, achieves optimal cooperation values in a small fraction of time compared with centralized algorithms, and adapts effectively to environmental changes. Keywords: MANET, trust models, game theory, evolutionary algorithm 1. Introduction Mobile Ad Hoc NETworks (MANETs) are infrastructureless networks formed by wireless mobile devices with limited resources. Source/destination pairs that are not within transmission range of each other must use intermediate nodes as relays (Perkins, 2001). The cooperation among nodes is fundamental for the operation of MANETs, since nodes that contribute with their own limited resources, such as battery, memory and processing capacity, should be able to use the resources contributed by other nodes. However, in this environment, there could be free- riders or selfish nodes, i.e., users that want to maximize their own welfare by using resources from the network to send their own packets without forwarding packets on behalf of others (Wrona and M¨ ah¨ onen, 2004). Thus, it is important to encourage nodes to participate in essential network functions such as packet routing and forwarding, because the higher the cooperation the better the network performance. In this sense, several trust mod els have been proposed as mechanisms to incentive node participation within the network (Mejia et al., 2009b). A trust * Corresponding author. Tel +573005645704. Email address: am.mejia75@uniandes.edu.co (Marcela Mejia) Preprint submitted to Journal of Network and Computer Applications (JNCA) September 4, 2010