Efficient and secure template blinding for biometric authentication.

Fingerprints as biometric authenticators are rapidly increasing in popularity, with fingerprint scanners available on many modern smartphones and laptops. Because these authenticators are non-revocable, special care must be taken to prevent leakage of the representative feature information of a user's fingerprint. While secure multiparty computation protocols have been designed to maintain fingerprint privacy during authentication, they do not protect the data stored on the authentication server. In this work, we develop a technique for blinding the stored biometric template such that the authentication server never observes biometric information in the clear, and cannot accidentally leak this information in the event of a breach. We show how our blinding technique can be combined with the privacy-preserving GSHADE protocol to privately compare biometric feature vectors using a variety of distance metrics with negligible overhead in computation time. We then construct a complete privacy-preserving remote fingerprint authentication system based on the Euclidean Distance metric, and show that a user can authenticate using privacy-preserving techniques in as little as 1.5 seconds. This work provides a template for designing low-cost blinding techniques for biometric authentication systems, and shows a practical use-case for secure multiparty computation protocols in remote authentication systems.