Mining approximate roles under important assignment

Abstract

Role Based Access Control (RBAC) has become the de facto access control model in recent years. In order to deploy RBAC, organizations have to define a set of roles from the existing user-permission assignment relationships, the process of which is called role mining. There have been many role mining algorithms proposed to devise a complete and correct set of roles which may not be very important because the user-permission assignment relationships are dynamic. This paper defines the δ-Approx Important Role Mining Problem (d-IRMP) which is proved to be NP-complete and proposes a heuristic bottom-up role mining algorithm that reduces the total number of roles with important assignments preserved at first. Furthermore, experiments were carried out with publicly available datasets to evaluate the approach and the results compared with other algorithms demonstrate the effectiveness of the proposed algorithm.