In magnetic flux leakage (MFL) detection, the identification of surface and back-side defects is required to obtain more accurate defect quantification and risk assessment results. However, current MFL techniques can detect both surface and back-side defects but are generally unable to distinguish between them. Therefore, this paper proposes a new boundary magnetic perturbation (BMP) testing method, combining the results of MFL to distinguish between surface and back-side defects. First, the detection mechanism of the BMP testing method and the impact of the tested magnetic flux density components are presented and analyzed by simulations to further develop an identification method. Then, the influences of the BMP sensor's lift-off and installation position are investigated by experiments to improve distinguishing performance. Finally, the repeated measurements show that the surface and back-side defects within the wide range of sizes can be identified accurately, even when the defect depths are in the range of 12.5%-87.5% of the sample thickness. Furthermore, the BMP testing method neither increases the length of the detection device nor requires additional magnetizers or signal generators. Therefore, the proposed method is highly suitable for the existing MFL detection devices to distinguish between surface and back-side defects.