隨著臉部辨識技術被廣泛地應用在生活中各式各樣的安全存取系統，人臉防偽之相關研究也受到了越來越多的關注。近年來，許多的人臉防偽方法在面對已知的偽造攻擊類別時，有很好的辨識能力，但是，當面對未曾出現過的偽造攻擊類型時，這些方法在辨識能力會出現大幅度的下降。科技的進步，讓偽造的攻擊類型不斷地推陳出新，為了避免模型在事先定義的偽造攻擊類別上過度擬合，我們需要讓模型在訓練時，學習到更具有泛化能力與辨別能力的特徵。在此篇論文中，我們提出一種拆分模型表徵的方法，透過兩階段式特徵學習，可以有效地將偽造相關的特徵與其他特徵分開來。拆分的偽造特徵可以增強人臉防偽任務中辨識未出現過之攻擊的能力。而不同於過去的方法使用單階段的設計，我們利用二階段的架構，能增加模型在訓練時的穩定度，並能提取更具備泛化能力的特徵，我們在許多跨攻擊類別之人臉防偽資料集上進行實驗，實驗結果顯示，我們的方法有著優異的辨識準確度，並能與其他頂尖的方法相比擬。

As face recognition is widely used in diverse security-critical applications, the study of face anti-spoofing (FAS) has attracted more and more attention. Several FAS methods have achieved promising performance if the attack types in the testing data are included in the training data, while the performance significantly degrades for unseen attack types. It is essential to learn more generalized and discriminative features to prevent overfitting to pre-defined spoof attack types. This paper proposes a novel dual-stage disentangled representation learning method that can efficiently untangle spoof-related features from irrelevant ones. Unlike previous FAS disentanglement works with one-stage architecture, we found that the dual-stage training design can improve the training stability and effectively encode the features to detect unseen attack types. Our experiments show that the proposed method provides superior accuracy than the state-of-the-art methods on several cross-type FAS benchmarks.

1 Introduction 1 1.1 Motivation............................... 1 1.2 ProblemStatement .......................... 2 1.3 Contributions ............................. 3 1.4 ThesisOrganization.......................... 4
2 Related Work 5 2.1 FaceAnti­Spoofing.......................... 5 2.2 Disentangledrepresentationlearning................. 5
3 Proposed Method 7
3.1 Live­infoFramework......................... 8
3.2 DisentanglementFramework..................... 8
3.3 SpoofCueModule .......................... 9
3.4 TrainingandTesting ......................... 12
3.5 ModelArchitecture.......................... 12
3.5.1 Thefirsttrainingstage .................... 12 3.5.2 Thesecondtrainingstage................... 13
4 Experiments 17
4.1 DatabasesandProtocols ....................... 17
4.2 EvaluationMetrics .......................... 17
4.3 ImplementationDetails........................ 17
4.3.1 TestingonSiW­M ...................... 18
4.3.2 Testing on MSU­MFSD, CASIA­MFSD, and Replay­Attack 19
4.3.3 TestingonSiW........................ 19
4.3.4 Cross­datasettesting ..................... 20
4.4 ComputingCost............................ 20 4.4.1 Comparisonwithothermethods ............... 20
5 Ablation study 26 5.1 Discussionofthedisentangledfeatures. . . . . . . . . . . . . . . . 26 5.2 ContributionofeachLossfunction.................. 28 5.3 VisualizationandAnalysis ...................... 29
6 Conclusions 34 References
35

[1] Arashloo, S. R., Kittler, J., and Christmas, W. An anomaly detection approach to face spoofing detection: A new formulation and evaluation protocol. IEEE access 5 (2017), 13868–13882.
[2] Biometrics., I. J. S. Information technology–biometric presentation attack detection–part 3: testing and reporting.
[3] Chingovska, I., Anjos, A., and Marcel, S. On the effectiveness of local binary patterns in face anti­spoofing. In 2012 BIOSIG­proceedings of the interna­ tional conference of biometrics special interest group (BIOSIG) (2012), IEEE, pp. 1–7.
[4] Da, K. A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014).
[5] deFreitasPereira,T.,Komulainen,J.,Anjos,A.,DeMartino,J.M.,Hadid,A., Pietikäinen, M., and Marcel, S. Face liveness detection using dynamic texture. EURASIP Journal on Image and Video Processing 2014, 1 (2014), 2.
[6] Feng, H., Hong, Z., Yue, H., Chen, Y., Wang, K., Han, J., Liu, J., and Ding, E. Learning generalized spoof cues for face anti­spoofing. arXiv preprint arXiv:2005.03922 (2020).
[7] George, A., and Marcel, S. Deep pixel­wise binary supervision for face pre­ sentation attack detection. In 2019 International Conference on Biometrics (ICB) (2019), IEEE, pp. 1–8.
[8] Huang, X., Liu, M.­Y., Belongie, S., and Kautz, J. Multimodal unsupervised image­to­image translation. In Proceedings of the European conference on computer vision (ECCV) (2018), pp. 172–189.
[9] Jourabloo, A., Liu, Y., and Liu, X. Face de­spoofing: Anti­spoofing via noise modeling. In Proceedings of the European Conference on Computer Vision (ECCV) (2018), pp. 290–306.
[10] Kim, T., Kim, Y., Kim, I., and Kim, D. Basn: Enriching feature representation using bipartite auxiliary supervisions for face anti­spoofing. In Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops (2019), pp. 0–0.
[11] King, D. E. Dlib­ml: A machine learning toolkit. The Journal of Machine Learning Research 10 (2009), 1755–1758.
[12] Kollreider, K., Fronthaler, H., Faraj, M. I., and Bigun, J. Real­time face de­ tection and motion analysis with application in “liveness”assessment. IEEE Transactions on Information Forensics and Security 2, 3 (2007), 548–558.
[13] Komulainen, J., Hadid, A., and Pietikäinen, M. Context based face anti­ spoofing. In 2013 IEEE Sixth International Conference on Biometrics: The­ ory, Applications and Systems (BTAS) (2013), IEEE, pp. 1–8.
[14] Lin, B., Li, X., Yu, Z., and Zhao, G. Face liveness detection by rppg features and contextual patch­based cnn. In Proceedings of the 2019 3rd International Conference on Biometric Engineering and Applications (2019), pp. 61–68.
[15] Liu, S., Yang, B., Yuen, P. C., and Zhao, G. A 3d mask face anti­spoofing database with real world variations. In Proceedings of the IEEE conference on computer vision and pattern recognition workshops (2016), pp. 100–106.
[16] Liu,S.­Q.,Lan,X.,andYuen,P.C.Remotephotoplethysmographycorrespon­ dence feature for 3d mask face presentation attack detection. In Proceedings of the European Conference on Computer Vision (ECCV) (2018), pp. 558–573.
[17] Liu, Y., Jourabloo, A., and Liu, X. Learning deep models for face anti­ spoofing: Binary or auxiliary supervision. In Proceedings of the IEEE confer­ ence on computer vision and pattern recognition (2018), pp. 389–398.
[18] Liu, Y., Stehouwer, J., Jourabloo, A., and Liu, X. Deep tree learning for zero­ shot face anti­spoofing. In Proceedings of the IEEE/CVF Conference on Com­ puter Vision and Pattern Recognition (2019), pp. 4680–4689.
[19] Liu,Y.,Stehouwer,J.,andLiu,X.Ondisentanglingspooftraceforgenericface anti­spoofing. In European Conference on Computer Vision (2020), Springer, pp. 406–422.
[20] Määttä, J., Hadid, A., and Pietikäinen, M. Face spoofing detection from single images using micro­texture analysis. In 2011 international joint conference on Biometrics (IJCB) (2011), IEEE, pp. 1–7.
[21] Pan, G., Sun, L., Wu, Z., and Lao, S. Eyeblink­based anti­spoofing in face recognition from a generic webcamera. In 2007 IEEE 11th International Con­ ference on Computer Vision (2007), IEEE, pp. 1–8.
[22] Paszke, A., Gross, S., Chintala, S., Chanan, G., Yang, E., DeVito, Z., Lin, Z., Desmaison, A., Antiga, L., and Lerer, A. Automatic differentiation in pytorch.
[23] Patel, K., Han, H., and Jain, A. K. Secure face unlock: Spoof detection on smartphones. IEEE transactions on information forensics and security 11, 10 (2016), 2268–2283.
[24] Qin, Y., Zhao, C., Zhu, X., Wang, Z., Yu, Z., Fu, T., Zhou, F., Shi, J., and Lei, Z. Learning meta model for zero­and few­shot face anti­spoofing. In Proceedings of the AAAI Conference on Artificial Intelligence (2020), vol. 34, pp. 11916–11923.
[25] Ronneberger, O., Fischer, P., and Brox, T. U­net: Convolutional networks for biomedical image segmentation. In International Conference on Medical im­ age computing and computer­assisted intervention (2015), Springer, pp. 234– 241.
[26] Van der Maaten, L., and Hinton, G. Visualizing data using t­sne. Journal of machine learning research 9, 11 (2008).
[27] Wang, G., Han, H., Shan, S., and Chen, X. Cross­domain face presentation attack detection via multi­domain disentangled representation learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2020), pp. 6678–6687.
[28] Wen, D., Han, H., and Jain, A. K. Face spoof detection with image distor­ tion analysis. IEEE Transactions on Information Forensics and Security 10, 4 (2015), 746–761.
[29] Xiao, T., Hong, J., and Ma, J. Elegant: Exchanging latent encodings with gan for transferring multiple face attributes. In Proceedings of the European conference on computer vision (ECCV) (2018), pp. 168–184.
[30] Xiong, F., and AbdAlmageed, W. Unknown presentation attack detection with face rgb images. In 2018 IEEE 9th International Conference on Biometrics Theory, Applications and Systems (BTAS) (2018), IEEE, pp. 1–9.
[31] Yang, X., Luo, W., Bao, L., Gao, Y., Gong, D., Zheng, S., Li, Z., and Liu, W. Face anti­spoofing: Model matters, so does data. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2019), pp. 3507–3516.
[32] Yu, Z., Li, X., Niu, X., Shi, J., and Zhao, G. Face anti­spoofing with hu­ man material perception. In European Conference on Computer Vision (2020), Springer, pp. 557–575.
[33] Yu, Z., Qin, Y., Li, X., Wang, Z., Zhao, C., Lei, Z., and Zhao, G. Multi­ modal face anti­spoofing based on central difference networks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (2020), pp. 650–651.
[34] Yu, Z., Wan, J., Qin, Y., Li, X., Li, S. Z., and Zhao, G. Nas­fas: Static­ dynamic central difference network search for face anti­spoofing. arXiv preprint arXiv:2011.02062 (2020).
[35] Zhang, K.­Y., Yao, T., Zhang, J., Tai, Y., Ding, S., Li, J., Huang, F., Song, H., and Ma, L. Face anti­spoofing via disentangled representation learning. In European Conference on Computer Vision (2020), Springer, pp. 641–657.
[36] Zhang, Z., Yan, J., Liu, S., Lei, Z., Yi, D., and Li, S. Z. A face antispoofing database with diverse attacks. In 2012 5th IAPR international conference on Biometrics (ICB) (2012), IEEE, pp. 26–31.
[37] Zinelabidine, B., Jukka, K., Li, L., Feng, X., and Hadid, A. Oulunpu: a mobile face presentation attack database with real­world variations. In Proc. IEEE Int. Conf. on Identity, Security and Behavior Analysis, ISBA (2017), pp. 1–7.