Can you distinguish noise from non-linear correlations? I feel like you can never prove the absence of non-linear correlation, but only showing it‘s presence on

Can you distinguish noise from non-linear correlations? I feel like you can never prove the absence of non-linear correlation, but only showing it‘s presence once modelled/detected