To maintain the integrity of information technology infrastructures within enterprises and organizations, it is imperative to implement robust and dependable access control mechanisms. A prevalent method is Role-Based Access Control (RBAC), wherein permissions are groped into roles, which are subsequently assigned to the users of an IT system. Such assignments are referred to as role concepts. The objective of the Role Mining Problem (RMP), whose decision variant is NP-complete, is to find a role concept with minimal number of roles and evolutionary algorithms have been applied as effective meta-heuristic techniques to approximate optimal solutions. Recent studies have demonstrated that the integration of expert knowledge through user interaction with running evolutionary algorithms can substantially enhance the optimization process. In this paper, expert knowledge is integrated into a evolutionary role mining algorithm by injecting favorable roles into the individuals representing role concepts. The impact of such role injections on the optimization progress is investigated and several survival strategies are presented to ensure that individuals with injected roles survive long enough to exert their beneficial effect on the optimization process. The proposed survival strategies are evaluated in a series of experiments.