Faced with the more insidious electricity stealing methods of high-frequency pulsed magnetic field, anti-theft detection based on magnetic field signals would fail due to noise. Aiming at this problem, a method combining SVD and wavelet transform is applied to anti-theft devices and proposed to remove noise and separate stolen electricity signal in detection. Firstly, the noisy magnetic field mixed signal is transformed by FFT to obtain the power spectrum, and the classical threshold is introduced to determine the number of narrow-band noises and the effective rank order of the SVD.Then the mixed signal is SVD decomposed, the singular value corresponding to the narrow-band noise is reset to zero, and then the signal is reconstructed. Finally, wavelet transform is used to remove the influence of random noise. Simulation and experimental results show that, this method can be effectively applied to the detection of electrical stealing. Compared with FFT threshold filtering method and EMD method, the suppression effect of the proposed method is more prominent.