Selfish and Malicious Behavior in Delay-Tolerant Networks Naércio Magaia, Paulo Rogério Pereira, Miguel P. Correia INESC-ID/IST/UTL, Rua Alves Redol, 9. 1000-029 LISBOA, Portugal {naercio.magaia, miguel.p.correia}@ist.utl.pt, prbp@inesc.pt Abstract: Delay-Tolerant Networks (DTNs) are composed of nodes that cooperate to forward messages despite connectivity issues. This paper focuses on the problem of some nodes making limited or no contribution to the network. Misbehaving nodes consume network resources, reducing its performance and availability, therefore they constitute an important problem that should be considered. We study the impact of node misbehavior on seven DTN routing protocols using a large set of simulations. The results show that different protocols are more resilient to different types of node misbehavior. Keywords: Delay-Tolerant Networks, Routing, Cooperation, Misbehavior. 1. Introduction Delay Tolerant Networks (DTNs) [1] are composed of nodes that cooperate with each other to forward data despite connectivity issues, e.g., long and variable delays, high error rates, and intermittent connectivity. Due to their characteristics, DTNs are not amenable to traditional routing protocols for Mobile Ad-Hoc Networks (MANETs), like AODV. An interesting case is Vehicular Delay Tolerant Networks (VDTNs) [2], in which vehicles communicate wirelessly with each other on a DTN manner to disseminate messages. Some potential applications are notification of traffic conditions, weather reports, advertisements, and web or email access. DTN routing protocols use a store, carry and forward approach, which implies some degree of cooperation among nodes, as nodes route other nodes’ messages, or pick them in one place and deliver them in another. In order to overcome the lack of end-to-end paths, the protocols replicate messages, if necessary, in each contact. The presence of misbehaving nodes, i.e., of nodes that do not follow the protocol –e.g., by using more than their share of resources – is a problem that has already been identified, but not thoroughly studied. One paper has shown that the performance of a DTN can be severely degraded if such nodes exist [3]. Nevertheless, misbehaving nodes are a real possibility. An important cause for misbehavior is selfishness: disseminating bogus delivery probability values in order to increase or decrease the probability of being chosen [3][4] or to save its own resources (storage, CPU, energy, etc.) [5]. Moreover, malicious attacks are pandemic in the Internet so they can also affect DTNs. The contribution of this paper is a study of the impact of misbehaving nodes on several representative DTN routing protocols in terms of the number of copies created – single- copy, n-copy, and unlimited-copy – and if an estimation metric is used – estimation-based routing protocols. We evaluated the DTN routing protocols’ performance in terms of delivery ratio, buffer time, hop count, latency and overhead ratio. Our work allows an adequate selection of DTN routing protocol if the presence of misbehaving nodes is possible or expected. However, we also conclude that there is no one-size-fits-all protocol,