Uncertainty assignment for measurement values is a common process for single sensors, but this procedure grows in complexity for sensor networks. Often measured values are processed further in such networks and uncertainty must be evaluated for virtual values. A simple example is the fusion of homogeneous values and faulty or drifting sensors can harm the virtual value. We introduce a method from the field of key-comparison into the domain of sensor fusion. The method is evaluated in three different scenarios within an agent-framework.