Counter-measures to photo attacks in face recognition: A public database and a baseline

By André Anjos, Sébastien Marcel. 2011

372 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Safety, Robustness & Interpretability, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Safety, Robustness & Interpretability

In Signal Processing

Citations in the corpus

Distillation as a Defense to Adversarial Perturbations Against Deep Neural Networks (2015, background): “Such attacks commonly occur today in non-DL classification systems [12], [13], [14], [15], [16].”
Statistical Meta-Analysis of Presentation Attacks for Secure Multibiometric Systems (2016, background): “) 60 Print Attack [26], [28] 50 12 16 3D Mask Attack [29] 17 5 (video) 10 (video)”
Deep Representations for Iris, Face, and Fingerprint Spoofing Detection (2014, methodology): “Both works validated the methods with the Print Attack benchmark [53].”
Face Anti-spoofing: Visual Approach (2014, methodology): “In [16], the authors made available a public dataset composed of printed photograph attacks and real-accesses, in which the samples available for the training and evaluating spoofing classifiers are videos.”
Face anti-spoofing using patch and depth-based CNNs (2017, background): “In order to develop a face recognition system that is invulnerable to various types of PAs, there is an increasing demand on designing a robust face anti-spoofing (or PA detection) system to classify a face sample as live or spoof before recognizing its identity [3, 4].”
LBP − T OP based countermeasure against face spooﬁng attacks (2012, background): “1 [1-2] [1-3] [1-4] [1-5] Rt 5 10 15 20 25 30 35”
Context based face anti-spoofing (2013, background): “If a background scene is incorporated with the face spoof (scenic fake face), overall motion correlation between the face and the background regions is likely to be observed [1, 27].”
Anti-spoofing in Action: Joint Operation with a Verification System (2013, background): “The third algorithm [3] analyzes the motion patterns present in the scene.”
Learning One Class Representations for Face Presentation Attack Detection Using Multi-Channel Convolutional Neural Networks (2020, methodology): “Often, features such as color, texture [5], [6], motion [7], and physiological cues [8], [9] and CNN”
Face Biometrics Under Spoofing Attacks: Vulnerabilities, Countermeasures, Open Issues, and Research Directions (2014, background): “In [2], the authors made available a public dataset composed of printed photograph attacks and real-accesses, in which the samples available for the training and evaluating spoofing classifiers are videos.”
Image Quality Assessment for Fake Biometric Detection: Application to Iris, Fingerprint, and Face Recognition (2014, background): “actions in modalities such as the iris [2], the fingerprint [17], the face [13], the signature [18], or even the gait [19] and multimodal approaches [20].”
Face Spoofing Detection Using Colour Texture Analysis (2016, background): “If a face spoof is not tightly cropped around the targeted face or it has an incorporated background scene (scenic fake face), it should be possible to observe high correlation between the overall motion of the face and the background regions for stationary face recognition systems [39], [40].”
Motion-based counter-measures to photo attacks in face recognition (2014, background): “c) Reference System 3 (RS3) - Correlation with simple frame differences: In this algorithm, extracted from [13], the movement direction is ignored and one focuses on intensity only.”
DeepFense: Online Accelerated Defense Against Adversarial Deep Learning (2017, background): “ reliable and trustworthy [1]. The existence of adversarial samples and their severe impact on the integrity of autonomous systems have been shown in the literature for both shallow [14], [40], [41], [42], [43] and deep [11], [12], [13], [44] learning models. Depending on the adversary goals, the ad”