In persasive environments, with the possibility of offering users distributed access on applications and services, from anywhere and at anytime, new issues arise with regard to access control mechanism. Generaly, the existing access control solutions make static user-permission associations and are unaware of the situation (context) when defining and enforcing access control policies. In order to address these issues, we propose a family of Context-Based Access Control models, named CxtBAC (Context-Based Access Control), which is composed by eight conceptual models that can be used as basis to construct context-based access control solutions. CxtBAC models explore contextual information as central concept for assigning permissions to users. In fact, context information can describe the situation of resource owners, resource requestors, resources, and the environment around them. Unlike existing access control proposals such as RBAC-based solutions (Role-Based Access Control), CxtBAC makes access decisions taking into account the contextual information that characterizes the situation of involved entities (e.g., resource owner and resource requestor). In a CxtBAC access rule, a set of permission is associated with an access context and users are dynamically associated with that access context. CxtBAC is independent of security policy language used to describe access control policies. As part of this work, we have proposed an implementation of CxtBAC policies based on ontologies and inference rules.