A New Protocol to Evaluate the Resistance of Template Update Systems against Zero-Effort Attacks

Since several years, there is a high interest in the creation of template update mechanisms in biometrics. By automatically updating over time the biometric reference of an individual, these update systems aim at avoiding a performance decreases over time. However, update errors occur when impostor samples are included in the biometric reference. In this paper, we are interested in analysing the behavior of template update against zero effort attacks depending on the difficulty of these attacks which is parametrized by the ratio of impostors samples included. We propose a protocol, which allows to evaluate an update mechanism under such kind of attacks by preserving an important amount of genuine samples, even when the ratio of impostor samples is important. Then, we apply this protocol on an existing template update system from the state of the art at various attack rates in order to assess its benefits.