Qualitative trust management methods

David Jelenc (2014) Qualitative trust management methods. PhD thesis.

Preview

PDF Download (1836Kb)

Abstract

In performing various tasks we often rely on others. While driving, we trust the manufacturer to produce a working car, civil engineers to build safe roads, and fellow drivers to comply to driving rules; if we did not, we would not be driving. Trust thus allows us to function in an unpredictable environment. For similar reasons, it is also becoming prevalent in computing, particularly in distributed systems, in which different users fulfill their goals by interacting with and thus relying on each other. Trust management systems are computer programs that advise users about trustworthiness of other entities in the system. This dissertation tackles two problems. First, we deal with trust models—mathematical procedures that compute trust—that use data presented with linguistic labels, which we term qualitative values. The model uses qualitative data to express assessments in past interactions, judgments in received opinions, and degrees of trust in trust estimations. We manipulate qualitative values appropriately by devising a method for estimating the most representative qualitative value in a collection of such values. We experimentally evaluate the model and compare it against the most influential works in the literature. In the second part, we deal with the evaluation of trust models. We propose a method that can directly evaluate and compare the performances of trust models regardless of their type: be it qualitative or quantitative. We implement the method and devise an experiment, which shows that the assumption underlying most of existing approaches, does not always hold. This shows that our approach, free of this assumption, is more general.

Actions (login required)