Privacy is universally regarded as an important consideration for pervasive computing environments. This is especially true when considering context-aware systems, because of the fact that context often contains personal information. Moreover, an adversary who takes the semantics of the context information into account can potentially infer knowledge beyond the information represented in the actual context data. In this paper, we describe a generic framework to model privacy in context-aware systems and discuss possible inferences based on the context of ID, location and time.