Abstract. E-government services usually process large amounts of confidential data. Therefore, security requirements for the communication between components have to be adhered in a strict way. Hence, it is of main interest that developers can analyze their modularized models of actual systems and that they can detect critical patterns. For this purpose, we present a general and formal framework for critical pattern detection and user-driven correction as well as possibilities for automatic analysis and verification at meta-model level. The technique is based on the formal theory of graph transformation, which we extend to transformations of type graphs with inheritance within a type graph hierarchy. We apply the framework to specify relevant security requirements. The extended theory is shown to fulfil the conditions of a weak adhesive HLR category allowing us to transfer analysis techniques and results r this abstract framework of graph transformation. In particular, we discuss ho...