You are here: Home Kolloquium Multimodal Biometric Presentation Attack Detection

Multimodal Biometric Presentation Attack Detection

Dr. Marta Gómez Barrero, Hochschule Darmstadt

Event details
What
    When Sep 13, 2019
    from 10:00 to 11:00
    Where Seminar-Raum U1.039 (Untergeschoss)
    Add event to calendar vCal
    iCal

    The deployment of biometric recognition systems has seen a considerable increase over the last decade, in particular for fingerprint based systems. To tackle the security issues derived from presentation attacks launched on the biometric capture device, automatic presentation attack detection (PAD) methods have been proposed. In spite of their high detection rates on the LivDet databases, the vast majority of the methods rely on the samples provided by traditional capture devices, which may fail to detect more sophisticated presentation attack instrument (PAI) species. In this paper, we propose a multi-modal fingerprint PAD which relies on an analysis of: i) the surface of the finger within the short wave infrared (SWIR) spectrum, and ii) the inside of the finger thanks to the laser speckle contrast imaging (LSCI) technology. On the experimental evaluation over a database comprising more than 4700 samples and 35 PAI species, and including unknown attacks to model a realistic scenario, a Detection Equal Error Rate (D-EER) of 0.2% has been achieved. Moreover, for a BPCER under 0.1% (i.e., highly convenient system), the APCER remains around 1%.
     

    Document Actions