Noise is a random phenomenon.
There are two particularly important types of noise that obey Poisson or Gauss statistics.
Noise should not be confused with interference or observational bias. The bias affecting a signal has properties that make it quite distinct from noise. One very common type of bias is offset, which is a shift in the signal due to the fact that the zero level of the physical signal and the zero level of the associated electrical signal are not the same.
The distinction between noise and an interference signal can be complex. We can take the example of the dark current of a detector which is superimposed on the observed signal. Its mean value is perfectly quantifiable (so many photoelectrons per pixel per second), which shows that although it is interference, it is indeed a signal with all the properties of a signal. The real noise component is in the fluctuations in the dark current.