Detecting face presentation attacks in mobile devices with a patch-based CNN and a sensor-aware loss function

Abstract

With the widespread use of biometric authentication comes the exploitation of presentation attacks, possibly undermining the effectiveness of these technologies in real-world setups. One example takes place when an impostor, aiming at unlocking someone else's smartphone, deceives the built-in face recognition system by presenting a printed image of the user. In this work, we study the problem of automatically detecting presentation attacks against face authentication methods, considering the use-case of fast device unlocking and hardware constraints of mobile devices. To enrich the understanding of how a purely software-based method can be used to tackle the problem, we present a solely data-driven approach trained with multi-resolution patches and a multi-objective loss function crafted specifically to the problem. We provide a careful analysis that considers several user-disjoint and cross-factor protocols, highlighting some of the problems with current datasets and approaches. Such analysis, besides demonstrating the competitive results yielded by the proposed method, provides a better conceptual understanding of the problem. To further enhance efficacy and discriminability, we propose a method that leverages the available gallery of user data in the device and adapts the method decision-making process to the user's and the device's own characteristics. Finally, we introduce a new presentation-attack dataset tailored to the mobile-device setup, with real-world variations in lighting, including outdoors and low-light sessions, in contrast to existing public datasets.

Facial expression recognition based on active region of interest using deep learning and parallelism

“…Since each region is processed individually, the results of these regions verify the accuracy of each other. This result confirms that the proposed technique achieves much-improved accuracy compared to earlier works ( Salmam, Madani & Kissi, 2016 ; Mollahosseini, Chan & Mahoor, 2016 ; AlBdairi, Zhu & Mohammed, 2020 ; Almeida et al, 2020 ; Rajan et al, 2020 ).…”

“…They used huge datasets in their experiment and applied GPUs technique to minimize the computation time. It is addressed in Geng et al (2019) that if ANN (artificial neural network) ( Suri, Vatsa & Singh, 2020 ) is applied with CNN ( Lucey et al, 2010 ; Mollahosseini, Chan & Mahoor, 2016 ; Yahia, Mohammad & Hossam, 2020 ; Almeida et al, 2020 ; Rajan et al, 2020 ) for detecting face then more accuracy can be gained. These detection methods are associated with correctness and recall rates, so they can be evaluated and applied in object detection.…”

“… Liu et al (2014) and Moghaddam, Akbarizadeh & Kaabi (2019) have tried to speed up the processing time by applying various techniques ( Almeida et al, 2020 ) in their research, but few of them became successful. Here processing time is reduced to a greater extent, although the measurement is done in floating-point, not in the exact value.…”

“…Classification of the desired expression: The prime goal of this paper remains in classifying the desired expressions out of five kinds (neutral, happiness, fear, surprise, and disgust). In this work, the Convolutional Neural Network (CNN) ( Ran et al, 2019 ; Galea & Farrugia, 2018 ; Jae & Bumshik, 2020 ; Ferrari et al, 2018 ; Sun et al, 2018 ; Mollahosseini, Chan & Mahoor, 2016 ; Shen, Wang & Liu, 2013 ; Yahia, Mohammad & Hossam, 2020 ; AlBdairi, Zhu & Mohammed, 2020 ; Almeida et al, 2020 ; Boyina, 2020 ; Rajan et al, 2020 ) is incorporated to classify the expression from the trained data. To classify the desired expressions, features are extracted first from the trained image sets.…”

Deep Learning for Face Anti-Spoofing: A Survey

“…For instance, high-fidelity PAs have similar appearance clues as the corresponding bonafide. On one hand, to learn a compact space with small intra-class distances but large inter-class distances, Hao [136] and Almeida et al [80] introduce contrastive loss and triplet loss, respectively. However, different from vision retrieval tasks, the bonafide and PAs in FAS task usually hold asymmetric intra-distributions (more compact and diverse, respectively).…”