Design and Performance Evaluation of Improved Genetic Algorithm for Role Mining Problem

Role Mining Problem (RMP) is an important issue in RBAC design and development. Genetic algorithm (GA) can be an effective method for solving RMP, but known usual GAs used for RMP have low performance at high dimensions. The paper proposes an improved GA for solving RMP. This algorithm is based on implementing some changes applied to the usual GAs. The main upgrades are the representation of algorithm chromosomes as strings of variable lengths with complex gene structures, the modernization of crossover operation, and the local optimization of chromosome structures after crossover execution on the basis of proposed rules. The performance evaluation results show that improved GA has better performance then usual GA. Moreover, the improved GA has a larger performance gain, when the required access control scheme is characterized by greater role severity.