Social caching and vehicular social networks: two contributions to the theory and practice of online social networks

Recent years have witnessed an explosive growth of the Online Social Networks (OSNs), which serve as a fertile ground for research such as, examining individual and group behaviors, identifying information dissemination patterns, and building unconventional OSNs. This dissertation explores the design and implementation of two emerging OSNs, a distributed social network (DOSN) and a vehicular social network (VSN), both from theoretical and practical perspectives, with the goal of improving the performance and functionality of the systems.Distributed OSNs have been proposed as an alternative to centralized OSNs. DOSNs neither have central repository of all user data, nor they impose control regarding how the data will be accessed, hence, enable users to keep control of their private data. One of the critical challenges in building DOSNs is the performance of the distributed social update dissemination. To address this, we propose Social Caches, nodes that cache updates of their friends in order to reduce the number of peer-to-peer network connections. We formulate social cache selection as an instance of the Neighbor-Dominating Set Problem, and propose a set of approximation, regression and heuristic social cache selection algorithms to solve it. Performance evaluation based on real social traffic data and five well-known social graphs shows that the proposed social caching mechanism can reduce P2P network connections by an order of magnitude.In the second part of the dissertation, we introduce a new class of online social networks called Vehicular Social Networks (VSNs), which connect drivers, commuters in particular, sharing the same road at the same time. As a first application over VSNs, we developed RoadSpeak, a scalable mobile Multiparty Voice Communication (MVC) system that allows drivers to communicate over automatically moderated chat groups based upon their interest. Each VSN chat group is regulated by a profile consisting of the triplet {time, location, interests }. Evaluation using benchmarks, field trials and simulations proves the feasibility of RoadSpeak, and demonstrates that it can support substantially larger groups of users compared with traditional MVC systems.