SaTC: CORE: Medium: Collaborative: Presentation-attack-robust biometrics systems via computational imaging of physiology and materials

SaTC：核心：中：协作：通过生理学和材料的计算成像实现演示攻击鲁棒生物识别系统

• 批准号: 1923184
• 负责人: Vishal Patel
• 金额: $ 30万
• 依托单位: Johns Hopkins University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2018
• 资助国家: 美国
• 起止时间: 2018-10-01 至 2022-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1923184&HistoricalAwards=false
• 关键词: SaTC; CORE; Medium; Collaborative; Presentation

Many physical characteristics, such as face, fingerprints, and iris as well as behavioral characteristics such as voice, gait, and keystroke dynamics, are believed to be unique to an individual. Hence, biometric analysis offers a reliable solution to the problem of identity verification. It is now widely acknowledged that biometric systems are vulnerable to manipulation where the true biometric is falsified using various attack strategies; such attacks are referred to as Presentation Attacks (PAs). This project develops computational imaging-based methods for detecting known and unknown PAs for face and iris biometric systems.While biometrics has been an active field of research for more than three decades, the design of biometric systems that are robust to a diverse set of known and unknown PAs is still in its infancy. This project seeks to build biometric systems that integrate computational imaging sensors for acquiring physiological, spectral, and material properties with the goal of detecting and mitigating the effects of known and unknown PA attacks. Specifically, as part of the project novel computational cameras will be built that augment traditional sensors used for face and iris recognition by adding subsystems that measure changes in a subject's physiology including vital signs, blood perfusion, voluntary and involuntary responses to external stimuli as well as composition as measured in scattering, reflectance and spectral profiles. The rich signatures sensed by these computational cameras is expected to provide a robust characterization of the true biometric and its variations while providing high discriminability to commonly used PAs. Further, by virtue of outlier detection and open-set modeling, this project develops a highly flexible approach for detecting previously unseen PAs.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

许多身体特征，如面部、指纹和虹膜，以及行为特征，如声音、步态和呼吸动力学，被认为是个体所独有的。因此，生物特征分析为身份验证问题提供了可靠的解决方案。现在人们普遍认为，生物识别系统容易受到操纵，其中使用各种攻击策略伪造真实的生物特征;这种攻击被称为呈现攻击（PA）。本项目开发了基于计算成像的方法，用于检测人脸和虹膜生物识别系统的已知和未知PA。虽然生物识别技术已经成为一个活跃的研究领域超过三十年，但对各种已知和未知PA具有鲁棒性的生物识别系统的设计仍处于起步阶段。该项目旨在构建生物识别系统，该系统集成了用于获取生理，光谱和材料特性的计算成像传感器，其目标是检测和减轻已知和未知PA攻击的影响。具体来说，作为该项目的一部分，将建造新型计算相机，通过增加测量受试者生理变化的子系统来增强用于面部和虹膜识别的传统传感器，包括生命体征，血液灌注，对外部刺激的自愿和非自愿反应以及散射，反射率和光谱分布测量的组成。由这些计算相机感测到的丰富签名预计将提供真实生物特征及其变化的鲁棒表征，同时为常用的PA提供高辨别力。此外，凭借离群值检测和开集建模，该项目开发了一种高度灵活的方法来检测以前看不见的PA。该奖项反映了NSF的法定使命，并被认为值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估来支持。

项目成果

A Joint Representation Learning and Feature Modeling Approach for One-class Recognition

• DOI: 10.1109/icpr48806.2021.9412390
• 发表时间: 2021-01
• 期刊: 2020 25th International Conference on Pattern Recognition (ICPR)
• 作者: Pramuditha Perera;Vishal M. Patel

One-Class Convolutional Neural Network

• DOI: 10.1109/lsp.2018.2889273
• 发表时间: 2019-02-01
• 期刊: IEEE SIGNAL PROCESSING LETTERS
• 影响因子: 3.9
• 作者: Oza, Poojan;Patel, Vishal M.
• 通讯作者: Patel, Vishal M.