From online social network analysis to a user-centric private sharing system

Online social networks (OSNs) have become a massive repository of data constructed from individuals’ inputs: posts, photos, feedbacks, locations, etc. By analyzing such data, meaningful knowledge is generated that can affect individuals’ beliefs, desires, happiness and choices—a data circulation started from individuals and ended in individuals! The OSN owners, as the one authority having full control over the stored data, make the data available for research, advertisement and other purposes. However, the individuals are missed in this circle while they generate the data and shape the OSN structure. In this thesis, we started by introducing approximation algorithms for finding the most influential individuals in a social graph and modeling the spread of information. To do so, we considered the communities of individuals that are shaped in a social graph. The social graph is extracted from the data stored and controlled centrally, which can cause privacy breaches and lead to individuals’ concerns. Therefore, we introduced UPSS: the user-centric private sharing system, in which the individuals are considered as the real data owners and provides secure and private data sharing on untrusted servers. The UPSS’s public API allows the application developers to implement applications as diverse as OSNs, document redaction systems with integrity properties, censorship-resistant systems, health care auditing systems, distributed version control systems with flexible access controls and a filesystem in userspace. Accessing users’ data is possible only with explicit user consent. We implemented the two later cases to show the applicability of UPSS. Supporting different storage models by UPSS enables us to have a local, remote and global filesystem in userspace with one unique core filesystem implementation and having it mounted with different block stores. By designing and implementing UPSS, we show that security and privacy can be addressed at the same time in the systems that need selective, secure and collaborative information sharing without requiring complete trust