Confidence Valuation in a Public-Key Infrastructure Based on Uncertain Evidence

Abstract

Abstract. Public-key authentication based on public-key certificates is a special case of the general problem of verifying a hypothesis (that a public key is authentic), given certain pieces of evidence. Beginning with PGP, several authors have pointed out that trust is often an uncertain piece of evidence and have proposed ad hoc methods, sometimes referred to as trust management, for dealing with this kind of uncertainty. These approaches can lead to counter-intuitive conclusions as is demonstrated with examples in the PGP trust management. For instance, an introducer marginally trusted by a user can make him accept an arbitrary key for any other user. In this paper we take a general approach to public-key authentication based on uncertain evidence, where not only trust, but also other pieces of evidence (e.g. entity authentication) can be uncertain. First, we formalize the assignment and the valuation of confidence values in the general context of reasoning based on uncertain evidence. Second, we propose a set of principles for sound confidence valuation. Third, we analyze PGP and some other previous methods for dealing with uncertainty in the light of our principles.

Efficient Distribution of Trust Authority Functions in Tactical Networks

“…The reduction to only four different types of trust allows the model to be implemented easily. However, Maurer [18] showed that due to this simplicity the model may delivers counter-intuitive results in special scenarios. A seminal approach to define a trust metric in the form of a model for public-key certification, trust and recommendations was defined by Maurer [3] in 1996.…”

On Data-Centric Trust Establishment in Ephemeral Ad Hoc Networks

“…In all traditional notions of trust, data trust (e.g., trust in the identity or access/attribute certificates) was based exclusively on a priori trust relations established with the network entities producing these data (e.g., certification authorities, network nodes) [9], [16], [17]. This was also the case when trust was derived via fairly lengthy interactions among nodes, as in reputation systems [4], [8], [18], [27].…”

“…Moreover, any new data trust relationships that needed to be established required only trust in the entity that produced those data. All trust establishment logics proposed to date have been based on entities (e.g., "principals" such as nodes) making statements on data [4], [7], [9], [12], [16], [17], [24], [25]. Furthermore, traditional trust relations evolved generally slowly with time:…”

A Trust Model for an Open, Decentralized Reputation System

“…A limit for the length of the trust chains can only be specified globally by the validator, but not by the issuer of the trust certificates. It has been shown in [17] that this model can produce counterintuitive results.…”

“…Several design principles and validation criteria for trust models have been proposed in [20,19] and [17], but there is no consensus on whether all trust models should follow these principles or whether trust models for different applications may have different requirements [21].…”