( Making templates compatible for template protection ) Biometric Feature-Type Transformation

IEEE SIGNAL PROCESSING MAGAZINE [77] SEPTEMBER 2015 B iometrics refers to physiological (i.e., face, fingerprint, hand geometry, etc.) and behavioral (i.e., speech, signature, keystroke, etc.) traits of a human identity. As these traits are unique to individuals, biometrics can be used to identify users reliably in many authentication applications, such as access control and e-commerce. Most biometric authentication systems offer great convenience without requiring the users to possess or remember any secret credentials. For applications that demand greater security, biometrics can be used in complement with passwords and security tokens to offer a multifactor authentication. Biometric measurement extracted from a user is subject to variations due to inconsistent environmental conditions or restrictiveness of the biometric representation used. The authentication decision of a biometric system is typically made upon how similar a sample is with reference to a template that is enrolled to the system at an earlier time. However, biometric templates used for similarity evaluation cannot be stored unprotected in the system because the consequences of biometric compromise is highly devastating. First, the stolen templates can be used to impersonate the corresponding users in the other applications. Second, biometrics is irrevocable and irreplaceable if it is compromised. A possible solution is to protect the biometric templates with a conventional encryption scheme such as the Rivest–Shamir–Adleman Digital Object Identifier 10.1109/MSP.2015.2423693