Cancellable biometrics and annotations on BioHash

摘要: Lately, the once powerful one-factor authentication which is based solely on either password, token or biometric approach, appears to be insufficient in addressing challenges of identity frauds. For example, sole approach suffers from privacy invasion and non-revocable issues. Passwords tokens are easily forgotten lost. To address these issues, notion cancellable biometrics was introduced denote templates that can cancelled replaced with inclusion another independent factor. BioHash a form mixes set user-specific random vectors features. In verification setting, able deliver extremely low error rates as compared when genuine used. However, this raises possibility two theft scenarios: (i) stolen-biometrics, an impostor possesses intercepted data sufficient high quality considered (ii) stolen-token, has access used by claim user. We found recognition rate for latter case poorer. paper, quantised projection ensemble Johnson-Lindenstrauss Lemma establish mathematical foundation BioHash. Based model, we elucidate characteristics pattern well security view points propose new methods rectify stolen-token problem.