The LivDet team asks to the participants to build a biometric system, composed of a "Comparator" module and a "Liveness" module, as in Figure 1, where the two modules cooperate in order to return a single output (IMSoutput). The goal is to simulate two scenarios, one in which a authorized person logs in using his fingerprint and another one in which an attacker attempts to evade the system. The circumvention of the system can be simulated in two ways through the fingerprint replica of an authorized person or through the attacker's live fingerprint. Therefore, in addition to the liveness metrics, in this LivDet edition we add two performance parameters:

  • Integrated Matcher Genuine accuracy (IMG_accuracy): the rate of correctly classified live fingerprints that belong to the people registered in the system.
  • Integrated Matcher Impostor accuracy (IMI_accuracy): the percentage of correctly classified fingerprints with which an attacker attempts to circumvent the system according to the second scenario.

To sum up, the parameters adopted for the performance evaluation are the following:

  •  Frej: Rate of failure to enroll.
  • IMG_accuracy: Rate of correctly classified genuine live fingerprints.
  • IMI_accuracy: Rate of correctly classified impostor live or genuine fake fingerprints.
  • Ferrlive: Rate of misclassified live fingerprints.
  • Ferrfake: Rate of misclassified fake fingerprints.

Ferrlive and Ferrfake are calculated from the livenessoutput; a livenessoutput between [0, 50) classify fingerprint image as "fake" while between [50,100] classify fingerprint image as "live".

IMG_accuracy and IMI_accuracy are calculated from IMSoutput (Integrated Match Score output): an IMSoutput between [0, 50) classify fingerprint image as "fake" or belonging to an attacker while between [50,100] classify fingerprint image as "live" and belonging to the declared user.

The winner will be awarded by simple averaging the overall classification errors on the datasets.

esempio funz2

Fig.1 Block diagram of a possible Integrated Match System. The livenessoutput is used to evaluate the Ferrlive and the Ferrfake. The IMSoutput is used to evaluate the IMG_accuracy and IMI_accuracy.