Guillermo Rodríguez Cano: Information Flow and Data Control in Online Social Networks: Insights and Lessons

Abstract 

Online social networks have become a practical and popular means for communication on the Internet, to some extent replacing traditional ones such as e-mail or bulletin boards. The free flow of large amounts of information, some of it of a personal and sensitive kind, has increased the need to know what happens with this data and how to achieve more control for the users.
We are interested in the relation between free flow of information and the desire to have control over data, for reasons such as privacy preservation or prevention of manipulation. We approach this relation from both sides, trying to understand how information flows and communication networks are formed as well as developing concrete mechanisms to control the information flow.
By understanding the characteristics of these flows in online social networks we learn that content is pushing the growth of online social networks and time is a driving factor for the evolution of these networks as well. For example, tight-knit short paths between users are analogous to cores of trust as the length of the path between users is a function of trust, or the distinctive ability of social structures to facilitate the rapid spread of information.
These characteristics reveal a multi-dimensional network whose abstract model is not simply a social graph. We find that relationships are of higher-order, comprising individuals and content, for which graphs need additional representation artefacts. We investigate the suitability of hyper- graphs as a tool for representation to mitigate semantic ambiguities and the potential loss of information for cases when pairwise relationships are superseded by the group.
On the side of controling data, we have worked on authenticity, accountability, and deniability aspects. Concretely, we developed protocols for usable password authentication in a decentralized manner as one building block for enabling increased user control over data by means of decentralized social networks. In this context, we also developed cooperation mechanisms without the need of a trusted third party for organizing events. Another mechanism was an anonymous document submission system with both unlinkability and provability for respective parties.
Our work in progress aims at combining these two approaches, understanding information flow and devising concrete mechanisms for data control in the case of disproportionate influence of users on a news web service based on community participation.