Simone Lupetti, Anders Andersen (Supervisor), Tage Stabell-Kulø (Supervisor), Information Protection - Information Sharing, Institutt for informatikk (trykkeri: Tromsprodukt), Tromsø, 2007, ISBN 978-82-92461-72-3.
Information is an important asset and as such, must be protected. On the other hand, except rare cases, the value of information increases when shared. This raises the paradox that the fewer are the obstacles hindering sharing, the more pressing are the needs for protection. Because sharing and protection are so intertwined, modern information systems must be designed to find the best possible trade-off between the two, for any given setting. A variety of techniques is available nowadays both to implement protection mechanisms and to enable efficient sharing but, the new applicative scenarios made possible by continuous technological advances require them to be updated or even redesigned from scratch. We start out by investigating the assumptions on which current protection mechanisms rest, as well as the consequences that their application can have on the host system. When using (the almost ubiquitous) public key cryptography, the trade-off between sharing and protection are deeply embedded in the system itself. If security is important at all, blocking (i.e. hindering sharing) sometimes becomes a necessity. We investigate the consequences of a security violation in public key systems from the security and availability point of view. Cryptographic protocols serve as another prominent tool that carries with it complex scenarios in which protection can be upheld only when the host system satisfies certain assumptions. We investigate the ramifications of using names and identifiers in cryptographic protocols. In the context of mechanisms for information protection, we also explore access control. We present a novel architecture designed for dynamic and multi-domain systems that maximizes sharing by enabling online delegation and decentralized access policies. We demonstrate also how physical proximity, when associated with cryptographic credentials (in the form of capabilities), can be successfully used to protect against unauthorized access and data aggregation attacks. In the second part of the thesis, we discuss some of the mechanisms designed to make possible or assist information sharing. In distributed systems, data is often replicated to increase performance and availability. We show how inconsistencies due to optimistic replication of the namespace (as opposed to conflicts in the data) can be resolved in a deterministic and user-friendly manner. Furthermore, to facilitate sharing, data must be distributed to the intended parties in the most efficient way. To this end, we investigate the transport order of data items over a single link when faced with restrictions on the available bandwidth, proposing a scheduling algorithm that minimizes the waiting time for a set of data consumers.