Role Mining in Access History Logs

Abstract

A novel approach for role mining in the context of role engineering for role-based access control is developed in this paper. We propose a simple algorithm, based on the assumption that permissions from the same role appear near each other in the access history log. Closely cooccurring groups of permissions are selected as candidate roles and are ranked based on a novel heuristic, called role cohesion, that quantizes the permission proximity of a candidate role in the access log. High-rank roles are identified using the algorithm, which is tested with a simulation scenario.