Safety of ABAC$_\alpha$ Is Decidable

Abstract

The $\mathrm {ABAC_\alpha }$ model was recently defined with the motivation to demonstrate a minimal set of capabilities for attribute-based access control (ABAC) which can configure typical forms of the three dominant traditional access control models: discretionary access control (DAC), mandatory access control (MAC) and role-based access control (RBAC). $\mathrm {ABAC_\alpha }$ showed that attributes can express identities (for DAC), security labels (for MAC) and roles (for RBAC). Safety analysis is a fundamental problem for any access control model. Recently, it has been shown that the pre-authorization usage control model with finite attribute domains ($\mathrm {UCON_{preA}^{finite}}$) has decidable safety. $\mathrm {ABAC_\alpha }$ is a pre-authorization model and requires finite attribute domains, but is otherwise quite different from $\mathrm {UCON_{preA}^{finite}}$. This paper gives a state-matching reduction from $\mathrm {ABAC_\alpha }$ to $\mathrm {UCON_{preA}^{finite}}$. The notion of state-matching reductions was defined by Tripunitara and Li, as reductions that preserve security properties including safety. It follows that safety of $\mathrm {ABAC_\alpha }$ is decidable.