Interoperability Among Capture Devices for Fingerprint Presentation Attacks Detection

Tuveri, Pierluigi; Ghiani, Luca; Zurutuza, Mikel; Mura, Valerio; Marcialis, Gian Luca

doi:10.1007/978-3-319-92627-8_4

Cited by 5 publications

(2 citation statements)

References 47 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Raja et al [164] concluded that handcrafted textural features achieve the best performance on capacitive sensors, whereas naturally learned features achieve optimal performance on thermal and optical sensors. They also suggested the use of deep learning-based approaches with the utilization of large data sets for the creation of new reliable PAD methods.…”

Section: Discussionmentioning

confidence: 99%

A comprehensive survey of fingerprint presentation attack detection

Κarampidis¹,

Rousouliotis²,

Linardos³

et al. 2021

JSSS

View full text Add to dashboard Cite

Section: Discussionmentioning

confidence: 99%

A comprehensive survey of fingerprint presentation attack detection

Κarampidis¹,

Rousouliotis²,

Linardos³

et al. 2021

JSSS

View full text Add to dashboard Cite

“…Figure 2 supports this idea, where we can observe the mean grayscale histogram for the five sensors investigated in this paper. Starting from this peculiarity, Tuveri et al [16] reached a significant level of interoperability by shifting sensor-specific feature distributions based on the least squares algorithm. The feature vectors were calculated by using textural algorithms.…”

Section: The Interoperability and Generalization Problemmentioning

confidence: 99%

Mitigating Sensor and Acquisition Method-Dependence of Fingerprint Presentation Attack Detection Systems by Exploiting Data from Multiple Devices

Micheletto

Orrù²,

Casula³

et al. 2022

Applied Sciences

View full text Add to dashboard Cite

The problem of interoperability is still open in fingerprint presentation attack detection (PAD) systems. This involves costs for designers and manufacturers who intend to change sensors of personal recognition systems or design multi-sensor systems, because they need to obtain sensor-specific spoofs and retrain the system. The solutions proposed in the state of the art to mitigate the problem still require data from the target sensor and are therefore not exempt from the problem of obtaining new data. In this paper, we provide insights for the design of PAD systems thanks to an overview of an interoperability analysis on modern systems: hand-crafted, deep-learning-based, and hybrid. We investigated realistic use cases to determine the pros and cons of training with data from multiple sensors compared to training with single sensor data, and drafted the main guidelines to follow for deciding the most convenient PAD design technique depending on the intended use of the fingerprint identification/authentication system.

show abstract