IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 15, NO. 1, FIRST QUARTER 2013 387 A Survey of Social-Based Routing in Delay Tolerant Networks: Positive and Negative Social Effects Ying Zhu, Bin Xu, Member, IEEE, Xinghua Shi, and Yu Wang, Senior Member, IEEE Abstract—Delay tolerant networks (DTNs) may lack continuous network connectivity. Routing in DTNs is thus challenging since it must handle network partitioning, long delays, and dynamic topology in such networks. In recent years, social-based ap- proaches, which attempt to exploit social behaviors of DTN nodes to make better routing decision, have drawn tremendous interests in DTN routing design. In this article, we summarize the social properties in DTNs, and provide a survey of recent social-based DTN routing approaches. To improve routing performance, these methods either take advantages of positive social characteristics such as community and friendship to assist packet forwarding or consider negative social characteristics such as selfishness. We conclude by discussing some open issues and challenges in social- based approaches regarding the design of DTN routing protocols. Index Terms—DTN routing; Social-based approaches; Social graphs; Social network analysis; Delay tolerant networks. I. I NTRODUCTION D ELAY or disruption tolerant networks (DTNs) [1]–[3] have recently drawn much attention from networking researchers due to the wide applications of these networks in challenging environments, such as space communications, military operations, and mobile sensor networks. Intermittent connectivity in DTNs results in the lack of instantaneous end- to-end paths, large transmission delay and unstable network topology. These characteristics make the classical ad hoc routing protocols [4]–[6] not being applicable for DTNs, since these protocols rely on establishment of a complete end-to-end route from the source to the destination. Many routing schemes [7]–[20] have been proposed for DTNs. Most of these DTN routing protocols belong to three categories: message-ferry-based, opportunity-based and prediction-based. In message-ferry-based methods [8]–[11], systems usually employ extra mobile nodes as ferries for message delivery. The trajectory of these ferries is controlled to improve delivery performance with store-and-carry. How- ever, controlling these nodes leads to extra cost and overhead. In opportunity-based schemes [3], [12], [13], nodes forward Manuscript received 4 January 2011; revised 10 June 2011 and 5 December 2011. Y. Zhu and Y. Wang are with Department of Computer Science, the University of North Carolina at Charlotte, 9201 University City Blvd., Charlotte, NC 28223, USA (e-mail: {yzhu17,yu.wang}@uncc.edu). B. Xu is with the Department of Computer Science and Technology, Tsinghua University, Beijing, 100084, China. X. Shi is with Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts 02215, USA. Digital Object Identifier 10.1109/SURV.2012.032612.00004 messages randomly hop by hop with the expectation of eventual delivery, but with no guarantees. Generally, messages are exchanged only when two nodes meet at the same place, and multiple copies of the same message are flooded in the network to increase the chance of delivery. Some DTN routing protocols [14]–[19] make relay selection by estimating metrics relative to successful delivery, such as delivery probability or expected delay based on a history of observations. Most of these protocols focus on whether two nodes will have a contact and when such contact happens if they do contact. Liu and Wu [20] also proposed a forwarding method based on a probabilistic forwarding metric, which is derived by modeling each forwarding as an optimal stopping rule problem. All of the current DTN routing methods share a similar paradigm, the “store and forward” fashion. If there is no connection available at a particular time, a DTN node can store and carry the data until it encounters other nodes. When the node has such a forwarding opportunity, all encountered nodes could be the candidates to relay the data. Thus, relay- ing selection and forwarding decision need to be made by the current node based on certain routing strategy. Various DTN routing approaches adopt different strategies based on different metrics. Example of such metrics include estimated delivery probability to the destination node, network resources available (including bandwidth, storage, and energy), esti- mated delay, and current network congestion level. However, the unpredictable mobility and restricted resource in DTNs significantly obstruct us from designing an ideal forwarding mechanism. Lately, the consideration of social characteristics provides a new angle of view in the design of DTN routing protocols. In most of the DTN applications (e.g. vehicular networks [21], [22], mobile social networks [23]–[27], disease epidemic spread monitoring and pocket switched networks (PSNs) [28]), a multitude of mobile devices are used and carried by people, whose behaviors are better described by social models. This opens new possibilities of social-based DTN routing, in which the knowledge of social characteristics are used to make better forwarding decision in DTN routing. Notice that social relations and behaviors among mobile users are usually long- term characteristics and less volatile than node mobility. Based on this observation and taking the recent advances in social network analysis, several social-based DTN routing methods [29]–[34] have been proposed recently to exploit various social characteristics in DTNs (such as community and centrality) 1553-877X/13/$31.00 c  2013 IEEE