Article
Logical Attacks and Countermeasures for Fingerprint
On-Card-Comparison Systems
Benoit Vibert , Jean-Marie Le Bars, Christophe Charrier and Christophe Rosenberger *
Ensicaen, Normandie University, Unicaen, CNRS, GREYC, 14000 Caen, France; benoit.vibert@ensicaen.fr (B.V.);
jean-marie.lebars@unicaen.fr (J.-M.L.B.); christophe.charrier@unicaen.fr (C.C.)
* Correspondence: christophe.rosenberger@ensicaen.fr
Received: 11 August 2020; Accepted: 11 September 2020; Published: 21 September 2020
Abstract:
Digital fingerprints are being used more and more to secure applications for logical and
physical access control. In order to guarantee security and privacy trends, a biometric system is often
implemented on a secure element to store the biometric reference template and for the matching with
a probe template (on-card-comparison). In order to assess the performance and robustness against
attacks of these systems, it is necessary to better understand which information could help an attacker
successfully impersonate a legitimate user. The first part of the paper details a new attack based on
the use of a priori information (such as the fingerprint classification, sensor type, image resolution
or number of minutiae in the biometric reference) that could be exploited by an attacker. In the
second part, a new countermeasure against brute force and zero effort attacks based on fingerprint
classification given a minutiae template is proposed. These two contributions show how fingerprint
classification could have an impact for attacks and countermeasures in embedded biometric systems.
Experiments show interesting results on significant fingerprint datasets.
Keywords: fingerprint classification; logical attack; evaluation; robustness; fingerprint features
1. Introduction
Biometrics is a commonly used technology for unlocking smartphones, secure border controls
or physical access to buildings. Yet, biometrics data are sensitive, since it is not possible in general to
revoke them in case of an attack. Thus, these data have to be protected as well as possible. In the case
of digital fingerprints, the reference template (a set of minutiae) is usually stored in a secure element
(SE) (such as e-passports). Due to the limitation of memory size and computational capabilities,
the reference template is stored following the ISO Compact Card II standard [
1
]. This representation
facilitates the comparison between the reference template and the probe sample. The security of
embedded biometric systems on a SE is therefore a primary requirement.
Regarding security, biometric systems have many vulnerabilities. As presented by Ratha et al. [
2
]
and more recently Jain et al. [
3
], authors have classified the attacks of a generic biometric system into
eight categories (as summarized in Figure 1). For each of the identified points, there are different types
of attacks. Uludag and Jain [
4
], Martinez [
5
] and Soutar [
6
] considered points 2 and 4 to perform a
hill-climbing attack. This attack can be performed by an application that continuously sends random
data to the system. The application retrieves the matching score between the reference template
and the probe sample and continues its disturbances only when the correspondence score increases
and until the acceptance threshold is reached. Note that on-card-comparison (OCC) systems never
provides as output the matching score in order to avoid this attack, and the decision is realized inside
the secure element.
In general, the attacker has to generate a biometric template to carry out an attack.
Considering embedded biometric comparison algorithms on a SE, the attacker sends random probes
Sensors 2020, 20, 5410; doi:10.3390/s20185410 www.mdpi.com/journal/sensors
