Design and analysis of social network systems (SNS)

Abbas, Syed Muhammad Ali (2016) Design and analysis of social network systems (SNS). Doctoral thesis (PhD), Manchester Metropolitan University.

Preview

Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (4MB) | Preview

Abstract

In the last few years, online Social Network Systems (SNSs) thrived and changed the overall outlook of the Internet. These systems play an important role in making the Internet social, a hallmark of Web 2.0. Various such systems have been developed to serve a diverse set of needs. SNSs provide not only a space for self-representation, but also mechanisms to build and maintain one’s social network online. A lot of studies have been carried out on such systems to identify how people develop cultures of communication, sharing and participation and also to identify the network structure of such systems. In this thesis, we carry this line of research forward. Our aim is the identification of some key user characteristics and social processes which result in the emergence of a social network. These might help future platform and application developers in creating better, more efficient and more open and user-friendly SNSs. Specifically, we make the following three major contributions: a) One of the distinct features of an SNS is the public listing of friendship links - social network. Most of the personal details such as hometown and workplace information have been hidden from non-friends, but the list of friendships remains open. Being a true representation, people use their real names as their screen names. Such names alone contain detailed cultural information about their ethnicities, religion and even their geographical origins. Our first contribution is that we have made good use of such information by inferring ethnic classification of users of Facebook. We identified how clustered and segregated the overall social network is when users’ inferred ethnicity is taken into account. Different cultures have different behaviours with distinct characteristics. This rich information can be used to develop an understanding and help create diverse applications catering for specific ethnicities and geographical regions; covering both the dominant and non-dominant groups. We have identified ethnicities of a subset of Facebook users with their friends and studied how different ethnicities are connected among and within each other. A large social network dataset of four thousand Manchester Metropolitan University (MMU) students have been selected from Facebook. We have extensively analysed this dataset for its network structure and also its semantic and social structure. Our work suggests our dataset is clustered and segregated on ethnic lines. b) To develop a user liberating SNS where the control and the ownership of rich personal data is in the hands of SNS users, a clear understanding is required of how such systems on an individual and group level are developed and maintained. Never before in Social Sciences was it possible to study society on such a large scale. These systems have facilitated the study of individuals both at a local and global scale. However, at the moment very little knowledge is available to identify how people develop their friendship in reality. So for example, it is not known whether in SNSs people meet others based on their attributes and interests, or if they simply bring online their real lives’ social networks. And more specifically, what processes does one go through to develop her social network. To fill this knowledge gap in this thesis, as our second contribution, we have used a computer simulation technique known as Agent-Based simulation, to develop four simulation models based on both individuals’ affinities and environmental aspects. Specifically, we have developed models of student interaction to develop social networks. Three University’s datasets which include Caltech (Nodes 762, Edges 16651), Princeton (Nodes 6575, Edges 293307) and Georgetown (Nodes 9388, Edges 425619), have been used to check the performance and rigour of the model. Our evidence suggests that ‘friend-of-a-friend’ (FOAF) best represents social interactions in Caltech University. In the case of Princeton and Georgetown, we found a multitude of social and structural processes involved, which are: attribute based (same dormitory, major or high school etc.), social interaction, random meet ups (through parties or other social events) and current friends introducing new friends. c) We observe that in the main, SNSs are centralised, and depend solely on central entities for everything. With huge personal data on such SNSs, advertising and marketing agencies have made very sophisticated systems to gather information about people. It is a goldmine for them for personalised advertisement. Also various governmental agencies have been using SNSs as an excuse to curb potential threats both legally and illegally, to obtain information on numerous users (people). In order to deal with such issues inherent in centralised client-server architecture, as the third contribution of this thesis, we have proposed and implemented a completely decentralised SNS in a peer-to-peer fashion. Our implementation is done in an open source Peer-To-Peer (P2P) client Tribler. To handle the dynamicity of users in a P2P system – their availability, we have developed mechanisms to deal with it. This SNS has been evaluated on a deployed system with real users. This prototype establishes the feasibility of a totally distributed SNS, but its practicality when scaled to a full system would require more work.