Multi-dimensional evidence-based trust management with multi-trusted paths

摘要

Trust management is an extensively investigated topic. A lot of trust models and systems have been proposed in the literature. However, a universally agreed trust model is rarely seen due to the fact that trust is essentially subjective and different people may have different views on it. We focus on the personalization of trust in order to catch this subjective nature of trust. We propose a multi-dimensional evidence-based trust management system with multi-trusted paths (MeTrust for short) to conduct trust computation on any arbitrarily complex trusted graph. The trust computation in MeTrust is conducted at three tiers, namely, the node tier, the path tier, and the graph tier. At the node tier, we consider multi-dimensional trust. Users can define a primary dimension and alternative dimensions on their own and users can make their own privileged strategies and setup weights for different dimensions for trust computation. At the path tier, we propose to use the Frank t-norm for users to control the decay rate for trust combination, which can be tuned in between the minimum trust combination (there is no decay in terms of the path length) and the product trust combination (the decay is too fast when the path length is relatively large). At the graph tier, we propose GraphReduce, GraphAdjust, and WeightedAverage algorithms to simplify any arbitrarily complex trusted graph. We employ trust truncation and trust equivalence to guarantee that every link in the graph will be used exactly once for trust computation. We evaluated trust truncation ratio and trust success ratio through extensive experiments, which can serve as a guide for users to select from a wide spectrum of trust parameters for trust computation.