On Maximising the Utility of Mobile Social Networks

Social network analysis is the study of social relationships between individuals in a society. In recent years the increasing amount of interactions between individuals using mobile devices has motivated research in social network analysis of mobile networks. Portable devices, such as blackberries, mobile phones, personal digital assistants have gained wide popularity. They are increasingly becoming part of our everyday life and are used pervasively. Thus the mobile software support a vast, diverse suite of applications from business and e-commerce trading to even dating and matchmaking services. One of the challenges in the area of mobile software is the amount of spam/unwanted context that users receive, as well as unsatisfactory and regretful interactions experienced by the users in different scenarios. In recent years, many researchers targeted this problem, however there has not yet been any work to take advantage of social analysis in designing a trust model. We believe that a trust model, founded on social network analysis, can help tackle these problems. For instance, by identifying the social structure of a network, such a trust model could gain intelligence about trustworthy agents, so that mobile software could be able to filter some of the irrelevant products or information (i.e. spam), or to restrict interactions to specific agents. The goal of this thesis is to combine social network analysis with distributed trust management in order to maximise the utility of a network. This utility could be defined, for example, as the number of good interactions/relevant context versus the number of malicious interactions/spam received. We will first demonstrate the usefulness of distributed trust management models in isolating misbehaving nodes. We will then analyse the social structure of networks, and demonstrate that agents do not exist in isolation but always as part of one (or more) community. Finally, we will combine social network analysis with distributed trust management and study the effect that such model has when computing the utility of an evolving network of agents.