Master Thesis: Creating Social Graph Overlays by Gossiping in Unstructured P2P Online Social Networks
Creating Social Graph Overlays by Gossiping in Unstructured P2P Online Social Networks
Recent developments in sensor technologies and web services make the smartphone the optimal social networking device: it typically has only one user, is highly personalized, and contains location and other context data about the user. Additionally to explicitly expressed friendship connections with other users, similar users – in terms of some context data, e.g., taste in music – can be found and connected to in a different layer of the social graph. To address privacy concerns, utilizing P2P technology for a distributed OSN, the social graph – indicating the connections between users – can be stored in a distributed manner.
In this work, the idea for an OSN in an unstructured P2P network is to utilize users’ smartphones as nodes that automatically collect location and context data to form a profile about its user. In order to connect to new users, generated profiles should be compared pairwise with other users. If the comparison indicates that two users are similar with respect to some profile feature, an edge in the social graph should be created. These edges can be named – after the profile feature that is similar – and weighted – with respect to the level of similarity. Such a social overlay could then be used for giving (friend) recommendations or disseminating content to relevant users. There is some related work in the field of comparing to profiles on the basis of Bloom filters. For unstructured P2P networks, gossip protocols are an established way of efficiently spreading information.
In this thesis, such a described unstructured P2P OSN should be simulated (e.g., with PeerSim1). This includes the creation of a dataset representing different users, the comparison of those users, and the gossiping of profile information in order to build and maintain social graphs. The configurable simulation should visualize or demonstrate the creation and change of those social graphs.
Supervisor: Felix Beierle, Sebastian Göndör 
Type: Master Thesis
Duration: 6 months
10587 Berlin, Germany
Phone: +49 30 8353 58811
Fax: +49 30 8353 58409