In this study of the face recognition on masked versus unmasked faces generated using Flickr-Faces-HQ and SpeakingFaces datasets, we report 36.78% degradation of recognition performance caused by the mask-wearing at the time of pandemics, in particular, in border checkpoint scenarios. We have achieved better performance and reduced the degradation to 1.79% using advanced deep learning approaches in the cross-spectral domain.