Progress in computer vision makes biometric authentication and recognition become more reliable and readily available than before, with a potential for ubiquitous use of such data well beyond traditional authentication. The need to protect highly sensitive biometrics is apparent, which, unlike other types of data used for authentication purposes, cannot be revoked and replaced with a new value. This means that any breach in security of biometric data has far reaching consequences to the individual than loss of other sensitive data.
Previously, only limited mechanisms for protecting privacy of biometry were available. To address the paucity of adequate protection mechanisms, this project develops novel techniques for enhancing privacy protection of biometric data when it is being processed and used in environments which are not fully trusted or secured in a broad set of contexts and applications that go well beyond comparing two biometric codes. The project's scope covers (i) secure interactive computation on biometric data, (ii) secure outsourcing of biometric computation to untrusted servers, and (iii) efficient verification of the correctness of the result of outsourced computation. A distinct benefit of this work is that it combines cryptographic design with practical evaluation on real biometric data, as well as includes integration of developed tools into a biometric computing grid environment. The outcomes of this research are expected to enable and promote safer practices in handing sensitive biometric data, including utilizing untrusted computing power, and a wider degree of collaboration between entities who are otherwise not permitted to share their data.