Data confidentiality and reputation schemes in distributed information systems

Abstract

In this thesis we discuss two demanding problems from the field of computer and communication security that involve trust. The first is known as the database service provider problem: A database owner wants a database service provider (DSP) to host her database. She only trusts this DSP to a limited extent, so she does not want to rely solely on contractual solutions. It is therefore necessary to enforce confidentiality of her data by technical means. The second problem concerns a (potentially very large) number of network nodes in a peer-to-peer (P2P) environment. Both problems are notoriously hard because, other than in traditional computer security problems, the adversary has a lot of control over the situation. The untrusted DSP needs to be able to process the data without learning anything about it, which seems to be a contradiction. In P2P applications it is desirable that nodes can join anonymously, but anonymity makes it easy to spread false reputation information. A node that enters a P2P application network for the first time needs to trust the claimed observations of other nodes, independent of the rate of malicious behaviour. Our findings are not perfect solutions, but nevertheless instructive in several ways: We propose relaxed, but still practically useful, notions of security for the DSP problem; we identify theoretical limitations of the DSP solution space; and we gradually reduce the impact of adversarial behaviour in P2P reputation systems using heuristic methods. As a side effect of our work, we present a special-purpose framework for simulation of P2P reputation systems that can be used to compare and fine-tune previous and upcoming work.