FaceFirst, an enterprise face recognition platform for retail and public safety, today announced Mask-ID, a new privacy feature that helps individuals maintain anonymity within face recognition systems. Mask-ID instantly blurs facial images without compromising match accuracy.
Mask-ID is the latest expression of FaceFirst’s Privacy by Design development methodology, which also includes role-based data access, purging facial data at regular intervals and providing enterprise-level data encryption.
“Mask-ID is a testament to our privacy principles,” said Peter Trepp, CEO of FaceFirst. “While our technology fully encrypts all data, our mission will always be to improve the privacy and security of biometric surveillance as a whole. With Mask-ID, FaceFirst has further enhanced protections against rogue operators and breaches, and we will continue to innovate and adapt to lead the industry in protecting individual privacy.”
Unlike traditional video surveillance systems, which store all visual images recorded within camera view, face recognition solutions only temporarily store video clips of anonymous visitors containing facial images, while matching them against facial images in a database. Biometric databases used for security typically contain facial templates of documented offenders or those with outstanding arrest warrants, while access control databases typically contain templates representing authorized employees or individuals.
“Mask-ID is an important step forward by FaceFirst to provide shoppers with greater privacy than traditional surveillance,” says Randy Salley, President of Salley LLC, a management consulting group, and former SVP of Retail Technology at Walmart. “With innovations like Mask-ID, shoppers can expect their anonymity to be protected with an enhanced security layer, while retailers are still empowered to instantly identify individuals who pose a potential threat to retail shrink, shoppers, or employees.”
Mask-ID is generally available to customers using the latest version of the FaceFirst platform.