计算机科学

计算机科学 ›› 2024, Vol. 51 ›› Issue (6A): 230400022-7.doi: 10.11896/jsjkx.230400022

• 信息安全 • 上一篇    下一篇

具有对抗鲁棒性的人脸活体检测方法

王春东, 李泉, 付浩然, 浩庆波   

  1. 天津理工大学计算机科学与工程学院 天津 300384
    天津理工大学“智能计算及软件新技术”天津市重点实验室 天津 300384
  • 发布日期:2024-06-06
  • 通讯作者: 李泉(shadow.lee20@qq.com)
  • 作者简介:(michael3769@163.com)
  • 基金资助:
    科技助力经济2020重点专项(SQ2020YFF0413781)

Face Anti-spoofing Method with Adversarial Robustness

WANG Chundong, LI Quan, FU Haoran, HAO Qingbo   

  1. School of Computer Science and Engineering,Tianjin University of Technology,Tianjin 300384,China
    Tianjin Key Laboratory of Intelligence Computing and Novel Software Technology,Tianjin University of Technology,Tianjin 300384,China
  • Published:2024-06-06
  • About author:WANG Chundong,born in 1969,Ph.D,professor,is a member of CCF(No.16230M).His main research interests include big data and smart computing security,network security situation awareness,etc.
    LI Quan,born in 1997,postgraduate.Her main research interests include face anti-spoofing,capsule networks,etc.
  • Supported by:
    Science and Technology for Economy 2020 Key Project of China(SQ2020YFF0413781).

PDF (PC)

282

摘要： 现有人脸活体检测方法在深度神经网络的支持下已获得优秀的检测能力,但面临对抗样本攻击时仍呈现脆弱性。针对此问题,引入胶囊网络(Capsule Network,CapsNet)提出一种具有对抗鲁棒性的人脸活体检测方法FAS-CapsNet:通过CapsNet及其图像重建机制保留特征间关联,过滤样本中的对抗扰动;根据皮肤与平面介质的反射性质差异,以Retinex算法增强图像光照特征,增大活体与非活体人脸类间距离的同时破坏对抗扰动模式,进而提升模型准确性与鲁棒性。在CASIA-SURF数据集上进行实验可知:FAS-CapsNet对正负样本的检测准确率为87.344%,对比模型中最高准确率为78.917%,说明FAS-CapsNet具备充分的常规活体检测能力。为进一步验证模型鲁棒性,基于CASIA-SURF测试集生成两种对抗样本数据集并进行实验:FAS-CapsNet在两数据集上的检测准确率分别为84.552%和79.042%,较常规检测准确率下降3.197%和9.505%;对比模型在两数据集上的最高准确率分别为74.938%和41.667%,较常规检测下降5.042%和47.201%。可见FAS-CapsNet受对抗扰动影响更小,具有显著的对抗鲁棒性优势。

关键词: 人脸活体检测, 对抗鲁棒性, 胶囊网络, Retinex, 对抗样本

Abstract: The existing face anti-spoofing methods based on deep neural networks perform excellently now,but they are absolute weak when facing adversarial examples.To solve the problem,capsule network(CapsNet) is introduced to propose an adversarial robust method called FAS-CapsNet.The capsule structure and reconstruction mechanism of CapsNet are utilized to retain the correlation between features and filter the adversarial perturbations in images.The Retinex algorithm is utilized to enhance illumination features which show the difference of reflection properties between skin and planar medium,increasing the between-class distance of living and spoof faces and destroying the very adversarial perturbation modes in images,improving the accuracy and robustness of FAS-CapsNet.Experiments on CASIA-SURF show that the spoofing detection accuracy of FAS-CapsNet is 87.344%,and the highest accuracy of comparison models is 78.917%,which demonstrates that FAS-CapsNet is capable to solve general face anti-spoofing problems.This paper further generates two adversarial datasets from CASIA-SURF validation set to verify the robustness of each model.The accuracy of FAS-CapsNet on the two datasets is 84.552% and 79.042% respectively,which decreases by 3.197% and 9.505% compared to the previous results.The highest accuracy of comparison models on adversarial datasets is 74.938% and 41.667% respectively,which is 5.042% and 47.201% lower than that of the conventional detection.It proves that FAS-CapsNet is significantly robust in adversarial attacks.

Key words: Face anti-spoofing, Adversarial robustness, CapsNet, Retinex, Adversarial examples

中图分类号: 

  • TP391
[1]YAO Y J,MA X C,GONG W,et al.In the era of face authenti-cation,is your“face” still safe[J].Policy Research & Exploration,2017(23):28-29.
[2]LIU S Y.Criminal regulation of cracking face recognition and authentication from the perspective of cooperative crime[J].Journal of Henan University of Economics and Law,2021,36(4):41-51.
[3]JIANG F L,LIU P C,ZHOU X D.A review on face anti-spoofing[J].Acta Automatica Sinica,2021,47(8):1799-1821.
[4]GALBALLY J,MARCEL S.Face anti-spoofing based on general image quality assessment[C]//2014 22nd International Conference on Pattern Recognition.IEEE,2014:1173-1178.
[5]MA Y K,WU L F,JIAN M,et al.Algorithm to generateadversarial examples for face-spoofing detection[J].Journal of Software,2019,30(2):469-480.
[6]ZHANG B,TONDI B,BARNI M.Adversarial examples for replay attacks against CNN-based face recognition with anti-spoofing capability[J].Computer Vision and Image Understanding,2020,197-198:102988.
[7]MIAO J.Risk challenge and regulatory research of face recognition facing “easy to crack”[J].Journal of Information Security Research,2021,7(10):984.
[8]SABOUR S,FROSST N,HINTON G E.Dynamic routing between capsules[C]//Proc.of 31st Conference on Neural Information Processing Systems(NIPS).MIT Press,2017:3859-3869.
[9]SINGH A K,JOSHI P,NANDI G C.Face recognition with liveness detection using eye and mouth movement[C]//2014 International Conference on Signal Propagation and Computer Technology(ICSPCT 2014).IEEE,2014:592-597.
[10]MA Y X,TAN L,DONG X,et al.Interactive livenessdetection algorithm for VTM[J].Computer Engineering,2019,45(3):256-261.
[11]ZHANG J,ZHANG N N.Research on interactive face detection based on optimized feature extraction[J].Computer Engineering and Applications,2019,55(13):193-200.
[12]MÄÄTTÄ J,HADID A,PIETIKÄINEN M.Face spoofing detection from single images using micro-texture analysis[C]//2011 International Joint Conference on Biometrics(IJCB).IEEE,2011:1-7.
[13]BOULKENAFET Z,KOMULAINEN J,HADID A.Face anti-spoofing based on color texture analysis[C]//2015 IEEE International Conference on Image Processing(ICIP).IEEE,2015:2636-2640.
[14]FENG J,DONG Z Y,LIU T T,et al.Face anti-spoofingliveness detection combining DQ_ CoALBP with LPQ descriptors[J].Computer Engineering and Applications,2022,58(14):134-143.
[15]ZHOU J,SHU K,ZHAO D,et al.Domain adaptation based person-specific face anti-spoofing using color texture features[C]//Proceedings of the 2020 5th International Conference on Machine Learning Technologies.2020:79-85.
[16]SHU X,TANG H,HUANG S.Face spoofing detection based on chromatic ED-LBP texture feature[J].Multimedia Systems,2021,27:161-176.
[17]ZHOU J,SHU K,LIU P,et al.Face anti-spoofing based on dynamic color texture analysis using local directional number pattern[C]//2020 25th International Conference on Pattern Recognition(ICPR).IEEE,2021:4221-4228.
[18]ZHANG L B,PENG F,QIN L,et al.Face spoofing detectionbased on color texture Markov feature and support vector machine recursive feature elimination[J].Journal of Visual Communication and Image Representation,2018,51:56-69.
[19]SHU X,TANG H,YANG X B,et al.Research on face anti-spoofing algorithm based on DQ_LBP[J].Journal of Computer Research and Development,2020,057():1508-1521.
[20]DANIEL N,ANITHA A.Texture and quality analysis for face spoofing detection[J].Computers & Electrical Engineering,2021,94(1):107293.
[21]CHANG H H,YEH C H.Face anti-spoofing detection based on multi-scale image quality assessment[J].Image and Vision Computing,2022,121:104428.
[22]WEN D,HAN H,JAIN A K.Face spoof detection with image distortion analysis[J].IEEE Transactions on Information Forensics & Security,2015,10(4):746-761.
[23]ZHANG S,WANG X,LIU A,et al.A dataset and benchmark for large-scale multi-modal face anti-spoofing[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2019:919-928.
[24]ZHANG P,ZOU F,WU Z,et al.FeatherNets:Convolutionalneural networks as light as feather for face anti-spoofing[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops(CVPRW).IEEE,2019.
[25]SHEN T,HUANG Y.FaceBagNet:Bag-of-local-features model for multi-modal face anti-spoofing[C]//The IEEE Conference on Computer Vision and Pattern Recognition Workshops(CVPRW 2019).IEEE,2019.
[26]PADNEVYCH R,SEMEDO D,CARMO D,et al.Improvingface liveness detection robustness with deep convolutional generative adversarial networks[C]//2022 30th European Signal Processing Conference(EUSIPCO).IEEE,2022.
[27]LI X,WU W,LI T,et al.Face face liveness detection based on parallel CNN[J].Journal of Physics:Conference Series,2020,1549(4):042069.
[28]SHEKHAR S,PATEL A,HALOI M,et al.An ensemble model for face liveness detection[EB/OL].(2022-01-19)[2023-02-17].https://arxiv.org/abs/2201.08901.
[29]YU Z,WAN J,QIN Y,et al.NAS-FAS:Static-dynamic central difference network search for face anti-spoofing[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2020,43(9):3005-3023.
[30]AKHTAR N,MIAN A.Threat of adversarial attacks on deep learning in computer vision:A survey[D].Ithaca:Cornell University,2018.
[31]FROSST N,SABOUR S,HINTON G.DARCCC:Detecting adversaries by reconstruction from class conditional capsules[EB/OL].(2018-11-16)[2023-02-17].https://arxiv.org/abs/1811.06969.
[32]QIN Y,FROSST N,SABOUR S,et al.Detecting and diagnosingadversarial images with class-conditional capsule reconstructions[C]//International Conference on Learning Representations.2020.
[33]QIN Y,FROSST N,RAFFEL C,et al.Deflecting adversarial attacks[EB/OL].(2020-02-18)[2023-02-17].https://arxiv.org/abs/2002.07405.
[34]LAND E.Lightness and retinex theory[J].Journal of the Optical Society of America,1971,61(1):1-11.
[35]HOWARD A G,ZHU M,CHEN B,et al.MobileNets:Efficient convolutional neural net-works for mobile vision applications[EB/OL].(2017-04-17)[2023-02-17].https://arxiv.org/abs/1704.04861.
[36]KURAKIN A,GOODFELLOW I,BENGIO S,et al.Adversarial machine learning at scale[EB/OL].(2017-02-11)[2023-02-17].https://arxiv.org/abs/1611.01236.
[37]SU J,VARGAS D V,KOUICHI S.One pixel attack for fooling deep neural networks[J].IEEE Transactions on Evolutionary Computation,2019,23(5):828-841.
[38]LUO W,LI Y,URTASUN R,et al.Under-standing the effective receptive field in deep convo-lutional neural networks[C]//Proceedings of the 30th International Conference on Neural Information Processing Systems.2016:4905-4913.
[39]YU Z,ZHAO C,WANG Z,et al.Searching central differenceconvolutional networks for face anti-spoofing[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2020:5295-5305.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市渝北区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发