Fairness in Biometrics: a figure of merit to assess biometric verification systems

Machine learning-based (ML) systems are being largely deployed since the last decade in a myriad of scenarios impacting several instances in our daily lives. With this vast sort of applications, aspects of fairness start to rise in the spotlight due to the social impact that this can get in minorities. In this work aspects of fairness in biometrics are addressed. First, we introduce the first figure of merit that is able to evaluate and compare fairness aspects between multiple biometric verification systems, the so-called Fairness Discrepancy Rate (FDR). A use case with two synthetic biometric systems is introduced and demonstrates the potential of this figure of merit in extreme cases of fair and unfair behavior. Second, a use case using face biometrics is presented where several systems are evaluated compared with this new figure of merit using three public datasets exploring gender and race demographics.

Scalable Probabilistic Models for Face and Speaker Recognition

Laurent El Shafey

In the biometrics community, face and speaker recognition are mature fields in which several systems have been proposed over the past twenty years. While existing systems perform well under controlled recording conditions, mismatch caused by the use of different sensors or a lack of cooperation from the subject still significantly affects performance, especially in challenging scenarios such as in forensics. Furthermore, existing methods suffer from scalability issues, which prevents them from taking advantage of increasingly large amounts of training data. This is otherwise a promising approach to improve accuracy in such challenging scenarios. In this thesis we address these problems of mismatch and complexity by developing scalable probabilistic models that we apply to face, speaker and bimodal recognition. Our contributions are four-fold. First, we propose a unified framework for session variability modeling techniques based on Gaussian mixture models (GMM), that encompasses inter-session variability (ISV) modeling, joint factor analysis (JFA) and total variability (TV) modeling. Second, we propose a novel exact and scalable formulation of probabilistic linear discriminant analysis (PLDA), which is a probabilistic and generative framework that models between-class and within-class variations. This formulation solves a major scalability issue, by improving both the time complexity of the training procedure from cubic to linear with respect to the number of samples per class, and the complexity of the scoring procedure. Furthermore, the implementations of all the proposed techniques are integrated into a novel collaborative open source software library called Bob 1 that enforces fair evaluations and encourages reproducible research. Fourth and finally, large-scale experiments are conducted with all of the above machine learning algorithms on several databases such as FRGC for face recognition, NIST SRE12 for speaker recognition and MOBIO for bimodal recognition, showing competitive performances.

EPFL2014

Multi-system biometric authentication

Norman Hoon Thian Poh

Verifying a person's identity claim by combining multiple biometric systems (fusion) is a promising solution to identity theft and automatic access control. This thesis contributes to the state-of-the-art of multimodal biometric fusion by improving the understanding of fusion and by enhancing fusion performance using information specific to a user. One problem to deal with at the score level fusion is to combine system outputs of different types. Two statistically sound representations of scores are probability and log-likelihood ratio (LLR). While they are equivalent in theory, LLR is much more useful in practice because its distribution can be approximated by a Gaussian distribution, which makes it useful to analyze the problem of fusion. Furthermore, its score statistics (mean and covariance) conditioned on the claimed user identity can be better exploited. Our first contribution is to estimate the fusion performance given the class-conditional score statistics and given a particular fusion operator/classifier. Thanks to the score statistics, we can predict fusion performance with reasonable accuracy, identify conditions which favor a particular fusion operator, study the joint phenomenon of combining system outputs with different degrees of strength and correlation and possibly correct the adverse effect of bias (due to the score-level mismatch between training and test sets) on fusion. While in practice the class-conditional Gaussian assumption is not always true, the estimated performance is found to be acceptable. Our second contribution is to exploit the user-specific prior knowledge by limiting the class-conditional Gaussian assumption to each user. We exploit this hypothesis in two strategies. In the first strategy, we combine a user-specific fusion classifier with a user-independent fusion classifier by means of two LLR scores, which are then weighted to obtain a single output. We show that combining both user-specific and user-independent LLR outputs always results in improved performance than using the better of the two. In the second strategy, we propose a statistic called the user-specific F-ratio, which measures the discriminative power of a given user based on the Gaussian assumption. Although similar class separability measures exist, e.g., the Fisher-ratio for a two-class problem and the d-prime statistic, F-ratio is more suitable because it is related to Equal Error Rate in a closed form. F-ratio is used in the following applications: a user-specific score normalization procedure, a user-specific criterion to rank users and a user-specific fusion operator that selectively considers a subset of systems for fusion. The resultant fusion operator leads to a statistically significantly increased performance with respect to the state-of-the-art fusion approaches. Even though the applications are different, the proposed methods share the following common advantages. Firstly, they are robust to deviation from the Gaussian assumption. Secondly, they are robust to few training data samples thanks to Bayesian adaptation. Finally, they consider both the client and impostor information simultaneously.

EPFL2007

Mobile Biometry (MOBIO) Face and Speaker Verification Evaluation

Sébastien Marcel

This paper evaluates the performance of face and speaker verification techniques in the context of a mobile environment. The mobile environment was chosen as it provides a realistic and challenging test-bed for biometric person verification techniques to operate. For instance the audio environment is quite noisy and there is limited control over the illumination conditions and the pose of the subject for the video. To conduct this evaluation, a part of a database captured during the ``Mobile Biometry'' (MOBIO) European Project was used. In total there were nine participants to the evaluation who submitted a face verification system and five participants who submitted speaker verification systems. The nine face verification systems all varied significantly in terms of both verification algorithms and face detection algorithms. Several systems used the OpenCV face detector while the better systems used proprietary software for the task of face detection. This ended up making the evaluation of verification algorithms challenging. The five speaker verification systems were based on one of two paradigms: a Gaussian Mixture Model (GMM) or Support Vector Machine (SVM) paradigm. In general the systems based on the SVM paradigm performed better than those based on the GMM paradigm.