Prevailing evaluations of fingerprint recognition systems have been performed as end-to-end black-box tests of fingerprint identification or authentication accuracy. However, performance of the end-to-end system is subject to errors arising in any of its constituent modules, including: fingerprint scanning, preprocessing, feature extraction, and matching. Conversely, white-box evaluations provide a more granular evaluation by studying the individual sub-components of a system. While a few studies have conducted stand-alone evaluations of the fingerprint reader and feature extraction modules of fingerprint recognition systems, little work has been devoted towards white-box evaluations of the fingerprint matching module. We report results of a controlled, white-box evaluation of one open-source and two commercial-off-the-shelf (COTS) minutiae-based matchers in terms of their robustness against controlled perturbations (random noise and non-linear distortions) introduced into the input minutiae feature sets. Our white-box evaluations reveal that the performance of fingerprint minutiae matchers are more susceptible to non-linear distortion and missing minutiae than spurious minutiae and small positional displacements of the minutiae locations.