VChain: Establishing Trust Based on Verifiable Credential Chains

Verifiable credentials (VCs) appear as a promising solution for holders to prove to verifiers the possession of certain desirable qualities or characteristics in a trustworthy and privacy-aware manner. Currently, a VC is linked to very specific properties for which it is issued and its trustworthiness is associated to its root of trust. However, determining the trustworthiness of a digital entity, i.e., a user, a microservice, etc., on arbitrary properties cannot be currently done based on their existing VCs, even if such a derivation could be trivial in the physical world; e.g., the certified access to an organization subdomain should provide access to the domain that includes it. In this paper, we define and experimentally analyze three different algorithms, namely GSP-X, CAM and LPwE, while employ VCs of multiple holders to infer the trustworthiness of a specific entity to possess specific properties of interest within a certain context, e.g., the likelihood of a microservice to meet advertised qualities for which it possesses no VCs. With synthetic datasets of credential chains of varying degrees of noise and hidden credentials from the ground truth, we experimentally establish that CAM and LPwE can accurately infer additional information about an entity's attributes or qualities even based on 50% and 60% of available ground truth, respectively. The algorithms, when properly parametrized, can be strict and accurate enough to perform even access control or authorization in various practical settings. Moreover, they are able to estimate a level of confidence on the possession of specific attributes based on the VCs available at the holder. Implementing one of the proposed algorithms within a blockchain smart contract can serve as a trust broker for assessing trust among microservices in open systems.