They are actually working with rats at this time. The first couple of years that compiled a database of rat-neurons in detail: Form and function. They do test the simulation extensively: Connecting electrodes to the synapses to check out what combination of input signals cause what output signals. After wards they look at one of the brains building blocks: The neuronal column: You assemble 10'000 neurons and do the same again: Feed it input and verify the output. If the simulation and the real thing gives the same result, then your simulation is ok, otherwise you go and tweak it until you get the same results.
I don't know how they go about Human brains, I'm sure they can not easily compare the simulation with the real thing. There are no volunteers to give op a bit of brain to feed the experiments :-).
They also are the main user of a BlueGene supercomputer at EPFL to run the simulations.
We'll see where they get over time. Henry Markram, the project leader is excellent, so I'm confident.