Error Analysis of Identification Information
Calculated From Fingerprints
Endre SELÉNYI, Zoltán HORNÁK
For the verification of digital signatures it is
indispensable that the signing person must be unambiguously identified.
Cryptography solves this problem by identifying the signing key, however it is
only assumed that just its legal owner possesses the secret key, hence current
implementations cannot prove that the owner used the appropriate key being
based on only property and optionally knowledge based user identification.
My work aims to enforce this correspondence by integrating
the owner’s biometric identification into the key preparation process
extracting information from the owner’s fingerprint, without what the secret
key that is stored encoded cannot be prepared for signing. This way the key can
only be restored for generating the digital signature if the owner is
identified successfully.
For this reason some kind of information should be
extracted from the fingerprint image. This information may also be used for
other purposes as well (e.g. biometric file encryption or as personal ID).
On the other hand biometric identification methods rather
aim the comparison of biometric samples than the calculation of any kind of
personal identification information from them that further could be used for
secret key encryption. Therefore the proposed algorithm will both suffer from
the decision errors of the biometric identification method itself and the
problems introduced by the approach of extracting information.
The article and the lecture aim the introduction of the
possible sources of decision errors using a multilevel approach, in which each
level can be evaluated separately, finally resulting cumulative error rates.
This error model also provides the ability to compare the possibilities of
successful information extraction to the acceptation of the actual sample in a
conventional biometric authentication system.