From online social network usage to the design of distributed online social networks

Abstract

Online social networks (OSNs) successfully claimed their place among the most popular Internet services. OSNs form online communities among people with common interests, activities, backgrounds, and/or friendships. They enable people to keep in touch with their friends, find people that they lost contact with and even find new friends based on shared affinities such as groups, hobbies, interests or overlaps in friendship circles. They also give people as well as companies a platform to adjust their self presentation according to the current environment. On the other hand they raise new issues that the overall society has to deal with, e.g., the phenomenon of cyber bullying, or privacy issues. Apart from the often repeated advice to simply not use any OSN, there have in the recent years been several attempts to solve the privacy issues by technical means. One approach are Distributed Online Social Networks (DOSN). In a Distributed OSN the underlying infrastructure is provided by the DOSN users, so that there is no single company controlling the entire system. This can be a set of distributed servers, e.g., as Diaspora or Friendica. Another approach is to base it on a Peer-to-Peer-infrastructure (P2P). DOSNs enable privacy preserving social networking mainly with two techniques: data is encrypted and stored in a distributed manner. The encryption ensures that data no matter where it is stored cannot be accessed without the permission of the user. The distributed storage helps to make the users independent of any single company and gives them control over their data. In this thesis we study the aspect of data availability in a P2P-based DOSN. We focus on an infrastructure that takes advantage of the naturally given relations between the users of a social network and evaluate the possible data availability of such a system. Since there is no P2P-based DOSN we can observe for this purpose we need to study data availability by simulating the behavior of our DOSN system. For this simulation we need a set of parameters that should be as realistic as possible. We derive such parameters from real OSNs. We first present our analysis of data we collected from Google+ during its transition from a beta test to the public release. Google+ is the latest attempt of one of the big players, Google, to propagate its own OSN. Our data includes the social graph of Google+ over 6 weeks and one set of public profile data. This gives us the opportunity to link social relations to other aspects regarding the users, such as their location, whom they follow, etc. We follow the growth of Google+ and the effects of the public release on the graph structure. Next, we focus on DOSN design, in particular, we present our approach to social data replication in such a system, and evaluate the resulting data availability. We assume that users in a social relation with other DOSN users have a natural interest to replicate their data. We use this behavior as a replication strategy to spread data of the DOSN users. To make our analyses more realistic, we base them on real OSNs, e.g., we derive the onand offline behavior of the users from Facebook data. We find that the availability of the content increases drastically when compared to the online time of the user, e.g., by a factor of more than 2 for 90% of the users.