while they sound impressively 'sciencey', it reminds me of many psychological concepts where they commit the nominal fallacy of naming something but not really explaining it.
There is a definition of sorts, but very few studies bother to explain it.
So to identify the default mode network, you start with a key spot, which is usually in the posterior cingulate (located towards the back of the brain just on the surface where the two hemispheres meet (there's one on each side). Then you plot changes in its activation levels over time, using fMRI, while the person is resting with eyes closed. Then you look for other structures whose changes seem to be "in synch" with the changes in your posterior cingulate "seed" area. That is, you look for correlations in activation across time. Those regions that correlate above a certain level (that you set beforehand) are considered part of the network.
It depends a bit on how where you set your correlation cut-off. But there've been a big bunch of studies now, and most agree the network includes posterior cingulate and surrounding cortcial areas (called retrosplenial cortex), the hippocampus and surrounding regions, the medial prefrontal cortex.
Then comes the messy part, where people start to make inferences about what this network is "doing" while you rest. Memory, visual imagery, thinking and evaluating self are all functions associated with various structures in this network. But they do a lot of other stuff too. This is where it gets all blurry.
In any case, the anatomy of the networks is fairly clear. The function of the networks is still, shall we say "a work in progress"