Systems and security products based on the RBAC model have been widely introduced to enterprises. Especially, the demands on enforcement of enterprise-level security policies and total identity management are rapidly growing. The RBAC model needs to be extended to deal with various circumstances of large enterprises, such as geographical distribution and heterogeneous environments including physical access control. In this paper, we introduce a new RBAC model, suitable for single sign-on systems. This model optimizes evaluation of rule-based RBAC so that total operation costs and productivity can be improved. Furthermore, to select most cost-effective RBAC extensions for enterprise-wide requirements, we propose a quantitative risk evaluation method based on fault trees. We construct fault trees having security violation and productivity loss as top events, and RBAC standard functions and security incidents as basic events. Probabilities of the top events are computed for given RBAC models and operation environments. We apply this method to a real enterprise system using the above RBAC extension and the proposed model realizes more safety and productivity over the base model.