Data mining for privacy in social networks

1 January, 2011 to 31 December, 2014


  • Bettina Berendt

Online social networking and content sharing has become part of everyday life, resulting in many people sharing the most intimate details of their personal life on social networking sites such as Facebook, Netlog or Twitter. While the amount of research on Social Network Analysis and privacy has grown substantially in the past years, many fundamental computational questions remain unanswered. Specifically: Social networks are generally modelled as graphs whose nodes represent natural persons, and whose edges represent friendship links. Privacy is assumed to be given when certain information is not accessible or can be transferred between
nodes without being revealed. This simple model is insufficient
conceptually and algorithmically, does not leverage the interdependencies between individuals as a core element of the Social Web, and does not take into account third party information.

This project aims to fill this gap. We will study: (1) privacy design types; (2) local-global dynamics; (3) intelligent micro-identity management; and (4) privacy for groups. We will
develop new techniques for (i) traffic analysis to analyze
social-network-based interactions and user behaviour; (ii) metrics to measure private information leakage; (iii) pattern mining to identify realistic outcomes, based on simulation models validated in the application area of Web-based social networks.

More details: here