Surveillance systems are already widely used at places such as automatic teller machines (ATMs), banks, and convenience stores; most of these surveillance systems record information about faces. However, criminals who illegally withdraw money from ATMs frequently cover their faces. In Taiwan it is common for criminals to cover their faces with safety helmets or masks. The authors previously proposed a technology for safety helmet detection; the current article presents a technology for the detection of masks that is based on automatic face recognition methods. The authors utilize Gabor filters to generate facial features which are transformed into binary-block images. The feature blocks are then analyzed using facial feature geometry analytic approaches to determine if the target person wears a mask. This type of analysis can give an early warning to possible problems when a mask is detected on a “customer” or intruder. Preliminary results are offered demonstrating the accuracy and reliability of the proposed technology. Figures, references