The operation of the power system is today highly dependent on computerized control systems. These SCADA systems resemble the central nervous system of the power system. At the same time as control systems enables more efficient, qualitative, and safe power systems, their vulnerabilities are also vulnerabilities to the power system. This paper presents a modeling language specifically developed for assessing the cyber security of SCADA systems. The modeling language uses the formalism Probabilistic Relational Models to integrate a mathematical inference engine with the modeling notation. If a SCADA system is modeled using this cyber security modeling language the cyber security of this SCADA system can be assessed probabilistically. Given a graphical description of a system, a quantitative analysis of threats is provided. This makes it possible to use the framework for evaluating the current solution as well as elaborate with what-if scenarios and the trade-offs between these. This cyber security modeling language could for example be used to model two control centers and the communication between them together with security mechanisms such as access control and communication protection The modeling language can also be used to describe a complete SCADA system and infer its security. The data associated with the probabilistic inference engine is only preliminary. In this paper we present a case study where cyber security modeling language has been applied to assess the security of a SCADA system. It is demonstrated how the modeling language can be applied and how a value for security can be inferred from architectural models (using the preliminary data). Future work will focus on the quantitative side of the modeling language. Probabilities will be elicited from literature, experiments, and field studies and through the opinion of domain experts. A tool is also being developed to support inference and analysis.