Touchless fingerprint recognition system has been proposed as a novel technology for overcoming several weaknesses of touch-based fingerprint recognition system. However, since there are still no researches for fake fingerprint detection in touchless fingerprint recognition system, the touchless system is still vulnerable to spoof attacks by artificial fingerprints. Therefore, we propose an multi-spectral illumination system for fake fingerprint detection in touchless fingerprint recognition. Our device includes green (525nm), yellow (590nm) and red (625nm) light sources for acquiring fingerprint images. From the fingerprint images acquired at the each of illuminations, we analyze reflectance properties of input finger and these features are used for discriminating between live and fake fingerprints. The experiments show that our system can provide a definite discrimination between live and fake finger.