Gian Luca Marcialis

Rapid and brief communication: A study on the performances of dynamic classifier selection based on local accuracy estimation

Dynamic classifier selection (DCS) plays a strategic role in the field of multiple classifier systems (MCS). This paper proposes a study on the performances of DCS by Local Accuracy estimation (DCS-LA). To this end, upper bounds against which the performances can be evaluated are proposed. The experimental results on five datasets clearly show the effectiveness of the selection methods based on local accuracy estimates.

Combining flat and structured representations for fingerprint classification with recursive neural networks and support vector machines

We present new fingerprint classification algorithms based on two machine learning approaches: support vector machines (SVMs) and recursive neural networks (RNNs). RNNs are trained on a structured representation of the fingerprint image. They are also used to extract a set of distributed features of the fingerprint which can be integrated in the SVM. SVMs are combined with a new error-correcting code scheme. This approach has two main advantages: (a) It can tolerate the presence of ambiguous fingerprint images in the training set and (b) it can effectively identify the most difficult fingerprint images in the test set. By rejecting these images the accuracy of the system improves significantly. We report experiments on the fingerprint database NIST-4. Our best classification accuracy is of 95.6 percent at 20 percent rejection rate and is obtained by training SVMs on both FingerCode and RNN-extracted features. This result indicates the benefit of integrating global and structured representations and suggests that SVMs are a promising approach for fingerprint classification.

Semi-supervised PCA-Based face recognition using self-training

Performances of face recognition systems based on principal component analysis can degrade quickly when input images exhibit substantial variations, due for example to changes in illumination or pose, compared to the templates collected during the enrolment stage. On the other hand, a lot of new unlabelled face images, which could be potentially used to update the templates and re-train the system, are made available during the system operation. In this paper a semi-supervised version, based on the self-training method, of the classical PCA-based face recognition algorithm is proposed to exploit unlabelled data for off-line updating of the eigenspace and the templates. Reported results show that the exploitation of unlabelled data by self-training can substantially improve the performances achieved with a small set of labelled training examples.

Security evaluation of biometric authentication systems under real spoofing attacks

Multimodal biometric systems are commonly believed to be more robust to spoofing attacks than unimodal systems, as they combine information coming from different biometric traits. Recent work has shown that multimodal systems can be misled by an impostor even by spoofing only one biometric trait. This result was obtained under a `worst-case` scenario, by assuming that the distribution of fake scores is identical to that of genuine scores (i.e. the attacker is assumed to be able to perfectly replicate a genuine biometric trait). This assumption also allows one to evaluate the robustness of score fusion rules against spoofing attacks, and to design robust fusion rules, without the need of actually fabricating spoofing attacks. However, whether and to what extent the `worst-case` scenario is representative of real spoofing attacks is still an open issue. In this study, we address this issue by an experimental investigation carried out on several data sets including real spoofing attacks, related to a multimodal verification system based on face and fingerprint biometrics. On the one hand, our results confirm that multimodal systems are vulnerable to attacks against a single biometric trait. On the other hand, they show that the `worst-case` scenario can be too pessimistic. This can lead to two conservative choices, if the `worst-case` assumption is used for designing a robust multimodal system. Therefore developing methods for evaluating the robustness of multimodal systems against spoofing attacks, and for designing robust ones, remain a very relevant open issue.

Fingerprint verification by fusion of optical and capacitive sensors

A few works have been presented so far on information fusion for fingerprint verification. None. however, have explicitly investigated the use of multi-sensor fusion, in other words, the integration of the information provided by multiple devices to capture fingerprint images. In this paper, a multi-sensor fingerprint verification system based on the fusion of optical and capacitive sensors is presented. Reported results show that such a multi-sensor system can perform better than traditional fingerprint matchers based on a single sensor.

Template Update Methods in Adaptive Biometric Systems: A Critical Review

Template representativeness is a fundamental problem in a biometric recognition system. The performance of the system degrades if the enrolled templates are un-representative of the substantial intra-class variations encountered in the input biometric samples. Recently, several template updates methods based on supervised and semi-supervised learning have been proposed in the literature with an aim to update the enrolled templates to the intra-class variations of the input data. However, the state of art related to template update is still in its infancy. This paper presents a critical review of the current approaches to template updating in order to analyze the state of the art in terms of advancement reached and open issues remain.

LivDet 2011 — Fingerprint liveness detection competition 2011

“Liveness detection”, a technique used to determine the vitality of a submitted biometric, has been implemented in fingerprint scanners in recent years. The goal for the LivDet 2011 competition is to compare software-based fingerprint liveness detection methodologies (Part 1), as well as fingerprint systems which incorporate liveness detection capabilities (Part 2), using a standardized testing protocol and large quantities of spoof and live fingerprint images. This competition was open to all academic and industrial institutions which have a solution for either software-based or system-based fingerprint vitality detection problem. Five submissions across the two parts of the competition resulted in successful completion. These submissions were: Chinese Academy of Sciences Institute of Automation (CASIA), Federico II University (Federico) and Dermalog Identification SystemsGmbH (Dermalog) for Part 1: Algorithms, and GreenBit and Dermalog for Part 2: Systems. Part 1 was evaluated using four different datasets. The best results were from Federico on the Digital Persona dataset with error for live and spoof detection of 6.2% and 11.61% respectively. The best overall results for Part 1 were Dermalog with 34.05 FerrFake and 11.825% FerrLive. Part 2 was evaluated using live subjects and spoof finger casts. The best results were from Dermalog with an error for live and spoof of 42.5% and 0.8%, respectively.

Fingerprint Liveness Detection using Binarized Statistical Image Features

Recent experiments, reported in the third edition of Fingerprint Liveness Detection competition (LivDet 2013), have clearly shown that fingerprint liveness detection is a very difficult and challenging task. Although the number of approaches is large, none of them can be claimed as able to detect liveness of fingerprint traits with an acceptable error rate. In our opinion, in order to investigate at which extent this error can be reduced, novel feature sets must be proposed, and, eventually, integrated with existing ones. In this paper, a novel fingerprint liveness descriptor named “BSIF” is described, which, similarly to Local Binary Pattern and Local Phase Quantization-based representations, encodes the local fingerprint texture on a feature vector. Experimental results on LivDet 2011 data sets appear to be encouraging and make this descriptor worth of further investigations.

Adaptive Biometric Systems That Can Improve with Use

Performances of biometric recognition systems can degrade quickly when the input biometric traits exhibit substantial variations compared to the templates collected during the enrolment stage of system’s users. On the other hand, a lot of new unlabelled biometric data, which could be exploited to adapt the system to input data variations, are made available during the system operation over the time. This chapter deals with adaptive biometric systems that can improve with use by exploiting unlabelled data. After a critical review of previous works on adaptive biometric systems, the use of semi-supervised learning methods for the development of adaptive biometric systems is discussed. Two examples of adaptive biometric recognition systems based on semi-supervised learning are presented along the chapter, and the concept of biometric co-training is introduced for the first time.

LivDet 2013 Fingerprint Liveness Detection Competition 2013

A spoof or fake is a counterfeit biometric that is used in an attempt to circumvent a biometric sensor Liveness detection distinguishes between live and fake biometric traits. Liveness detection is based on the principle that additional information can be garnered above and beyond the data procured by a standard verification system, and this additional data can be used to verify if a biometric measure is authentic. The Fingerprint Liveness Detection Competition (LivDet) goal is to compare both software-based (Part 1) and hardware-based (Part 2) fingerprint liveness detection methodologies and is open to all academic and industrial institutions. Submissions for the third edition were much more than in the previous editions of LivDet demonstrating a growing interest in the area. We had nine participants (with eleven algorithms) for Part 1 and two submissions for Part 2.