Whenever an abstract uses the word "noise" it has one of two meanings:

* An unpredictable component of a signal, due to things that are inherently theoretically unpredictable, such as Johnson–Nyquist noise.

* Part of a signal that the writers do not understand or cannot model due to shortcomings of the model. Describing such signals as 'noise' generally indicates a poor or incomplete paper.

I suspect in this case, it's the latter sadly.