Preference-aware object retrieval in opportunistic mobile social networks

Mobile social networks (MSNs) have attracted much attention for the past few years. Due to non-deterministic social contact and intermittent as well as asymmetric connectivity in MSNs, recent research works on MSNs only focus on unicast or multicast protocols for speeding up object dissemination. Unlike previous works, the focus of this paper is on how to design an efficient (multimedia) object retrieval scheme for MSNs. The key challenge for the object retrieval in MSNs is that we need to deal with not only the connectivity issue in MSNs, but also the heterogeneous objects owned by different users. Our idea is that, given a query object, we jointly consider the user preference probability and the characteristics of the round-trip object retrieval to design the object retrieval scheme. Since multimedia objects are usually associated with multiple attributes, we are able to characterize users by the objects they hold, and then compute the preference probability of the targeted object for each user. Next, we divide the object retrieval scenario in MSNs into two phases: the searching phase and the returning phase. When the query is in searching phase, the relaying user needs to jointly consider the successful probability in the searching phase as well as in the returning phase. In the returning phase, the relaying user just considers the probability of returning the object back to source node. In addition, we also develop a discrete-time analytic model to predict the node delivery probability and use it for the relay node selection. A performance study, via simulations based on MIT Reality and Last.fm music trace, illustrates that our proposed schemes are effective in object retrieval on MSNs.