Abstract <p>A method for statistical detection of weak ultra-low-frequency (ULF) electromagnetic anomalies of lithospheric origin amid intense anthropogenic noise is proposed. The method is based on the hypothesis that the statistical distribution of ambient electromagnetic noise amplitudes in the ULF range is stationary over time, whereas an additional signal perturbs this distribution. To validate this method, we conducted a numerical experiment with superimposing a synthetic noise signal onto the magnetic data recorded over a 72-day interval from December 2022 to February 2023 at the Tashkent Geodynamic Polygon. The results demonstrate that a persistent shift in the probability distribution mode of the root-mean-square deviations toward larger values serves as a reliable indicator of an electromagnetic disturbance from a remote source. The method was tested and its applicability was demonstrated for an <i>M</i>6.4 earthquake.</p>