Towards Biometric Digital Signatures
Orvos Péter <orvos@mit.bme.hu>
BME - Méréstechnika és Információs Rendszerek Tsz.
Selényi Endre Dr. <selenyi@mit.bme.hu>
BME - Méréstechnika és Információs Rendszerek Tsz.
Hornák Zoltán <hornak@mit.bme.hu>
BME - Méréstechnika és Információs Rendszerek Tsz.
The aim of digital signatures is to ensure the integrity of the signed document and to authenticate the signing person himself. The present implementations identify the signing person via the used secret key, assuming that nobody else can use the secret key. The association between the secret key and its owner is the weakest link in the chain of the PKI architecture; hence the enforcement of this link is inevitable for further increasing the level of security. A possible solution is, if some master secret is securely stored in an intelligent chip-card that must be combined with a biometric sample of the owner in order to receive the appropriate signing key. This way a successfully verifiable signature directly proves that the owner of the key was physically present when the signature was created, and he actively contributed in this act.
I analysed a method that derives the data vector, which can be used for the restoration of the signing key from the fingerprint image, precisely the positions of the minutiae points. In order to tune the algorithms parameters I have run several test calculations, during which I measured the extractable amount of information and the ratio of the two types of failure:
FAR: (False Acceptance Ratio) the algorithm makes a mistake if it restores the proper key from the fingerprint of someone else. Of course this event cannot be eliminated, since it would not be rational to assign unique data vector for every person of the world, but it is important to know how big user population can be expected, in which the repetition of this identification codes may occur.
FRR: (False Rejection Ratio) the algorithm also makes mistake if it calculates an improper personal vector from the fingerprint of the legal owner, and therefore creates the signature with a wrong key. This kind of failures can be noticed verifying the recently created signatures, but it is important from the usability point of view that failed trials in normal circumstances should occur very rarely.
In my lecture I will describe the analysed algorithm and the measurements I have run. I will talk about the aims of the tests, and finally I will introduce the measurement results and the conclusions that can be derived from them. I will close my lecture describing the directions, in which this algorithm can be made more suitable, and also introduce some new, alternative ideas.
As my lecture will prove, there are still several problems to be solved developing a biometric digital signature technology, I hope my results will convince everyone that secure and easy to use implementations will soon follow this idea.