Role Mining Algorithm Research Based On Cardinality Constraints

One key issue of the role-based access control (RBAC) is to build the role state. Al-though the way of building the role state has been a lot of research, these ways of buildingrole state didn’t only take the constraint information into consideration in the role-basedaccess control model, but don’t the assessment of the accuracy for the derived role statefrom the role mining algorithm. They just compare the least cost of the authorizationprocess; however, considering the constraint information is required by the role-basedaccess control model and is also an important motivation for RBAC model proposed.Atlast, role state accuracy assessment is required by the practical application.Therefore, it makes important value and significance for considering the constraint in-formation in the way of building the role state and making the assessment of the accuracyfor the derived role state from the role mining algorithm. Because four cardinality con-straints are pairwise mutually exclusive, we put forward a framework and take pairwisecardinality constraints in the framework. We propose role mining algorithms based on twoquarters cardinality constraints. The algorithm first creates the role set then updates therole set according to the algorithm based on Graph Optimization. For the previous algo-rithm, they don’t consider the similarity between roles so that the accuracy of the role stateis low. We correct the drawback by considering the similarity between roles in the processof merging roles.The experiments shows that although the algorithm considering permission cardinalityconstraint and user cardinality constraint increase five percent role numbers, the assign-ment cost reduces50percent.The experiments also shows that the algorithm consideringrole-permission cardinality constraint and role-user cardinality constraint doesn’t take thecost of the accuracy and cost. Instead, our algorithm still increases by20percent on theaccuracy when our algorithm spends the equivalent cost.