Paper 2010/246 Original: eprint.iacr.org/2010/246 ↗

Quantifying Trust

Abstract

Trust is a central concept in public-key cryptography infrastruc- ture and in security in general. We study its initial quantification and its spread patterns. There is empirical evidence that in trust-based reputation model for virtual communities, it pays to restrict the clusters of agents to small sets with high mutual trust. We propose and motivate a mathematical model, where this phenomenon emerges naturally. In our model, we separate trust values from their weights. We motivate this separation using real examples, and show that in this model, trust converges to the extremes, agreeing with and accentuating the observed phenomenon. Specifically, in our model, cliques of agents of maximal mutual trust are formed, and the trust between any two agents that do not maximally trust each other, converges to zero. We offer initial practical relaxations to the model that preserve some of the theoretical flavor.