This paper points out the importance of anti-aliasing filtering in the analog-to-digital conversion performed in electrochemical noise (EN) measurements. This operation is currently implemented in spectrum analyzers, but it is generally ignored in data acquisition systems based on acquisition cards in personal computers or on digital voltmeters. It is shown that without filtering, aliasing occurs and lends to fallacious amplitudes of the EN, that is, too-high standard deviations in the time domain and too-high power spectral density levels in the frequency domain. In particular the white noises, which are commonly encountered in localized corrosion at low sampling rates, always have a wrong amplitude when measured without filtering. Anti-aliasing filtering may be carried out with analog filters or a combination of analog and digital filters. A comparison of two different digital filters, which are easily implementable in personal computers, shows that the slope of the filter roll-off has to be high enough to avoid aliasing for all types of EN and to obtain a good overlapping of the power spectral densities measured at various sampling rates.