Existing security models for RDF use RDF patterns for defining the security policy. This approach leads to a number of security rules which rapidly tends to be unmanageable. In this paper we define a new security model which follows the traditional approach of creating security views, which has long been used by SQL database administrators. Our model first logically distributes RDF data into SPARQL views and then it defines security rules regulating SPARQL access to views. Moreover our model supports rights delegation and dynamic security rules (i.e. rules which can be active or not, depending on the context).