《繁凡的论文精读》（一）CVPR 2019 基于决策的高效人脸识别黑盒对抗攻击（清华朱军）-精品文档资料整理.docx

《《繁凡的论文精读》（一）CVPR 2019 基于决策的高效人脸识别黑盒对抗攻击（清华朱军）-精品文档资料整理.docx》由会员分享，可在线阅读，更多相关《《繁凡的论文精读》（一）CVPR 2019 基于决策的高效人脸识别黑盒对抗攻击（清华朱军）-精品文档资料整理.docx（4页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、繁凡的论文精读（一）CVPR 2019 基于决策的高效人脸识别黑盒对抗攻击（清华朱军）图6。对真实世界人脸验证API的模拟攻击的例子。我们展示了原始图像对和由每种方法产生的对抗图像。 5. Conclusion 在本文中 我们提出了一种进化攻击算法 用于在基于决策的黑盒环境中生成对抗实例。我们的方法通过对搜索方向的部分几何形状进展建模 同时降低搜索空间的维数 进而进步了效率。我们应用提出的方法综合研究了几种先进的人脸识别模型的鲁棒性 并与其他方法进展了比拟。大量实验证明了该方法的有效性。我们说明 现有的人脸识别模型极易受到黑盒方式的攻击 这为开发更鲁棒的人脸识别模型提出了平安问题。最后 利用该

2、方法攻击了一个真实世界的人脸识别系统 验证了其实用性。 References 1 W. Brendel, J. Rauber, and M. Bethge. Decision-based adversarial attacks: Reliable attacks against black-box machine learning models. In ICLR, 2018. 2, 5, 6, 8 2 N. Carlini and D. Wagner. Towards evaluating the robustness of neural networks. In IEEE Symposium

3、 on Security and Privacy, 2017. 2 3 P.-Y. Chen, H. Zhang, Y. Sharma, J. Yi, and C.-J. Hsieh. Zoo: Zeroth order optimization based black-box attacks to deep neural networks without training substitute models. In Proceedings of the 10th ACM Workshop on Articial Intelligence and Security, pages 1526. A

4、CM, 2017. 2, 3, 5 4 M. Cheng, T. Le, P.-Y. Chen, J. Yi, H. Zhang, and C.-J. Hsieh. Query-efcient hard-label black-box attack: An optimization-based approach. arXiv preprint arXiv:1807.04457, 2018. 2, 3, 5, 6, 8 5 J. Deng, J. Guo, and S. Zafeiriou. Arcface: Additive angular margin loss for deep face

5、recognition. arXiv preprint arXiv:1801.07698, 2018. 1, 2, 5, 6, 7 6 Y. Dong, F. Liao, T. Pang, H. Su, J. Zhu, X. Hu, and J. Li. Boosting adversarial attacks with momentum. In CVPR, 2018. 1, 2 7 A. D. Flaxman, A. T. Kalai, and H. B. Mcmahan. Online convex optimization in the bandit setting:gradient d

6、escent without a gradient. In Sixteenth ACM-SIAM Symposium on Discrete Algorithms, pages 385394, 2005. 3 8 S. Ghadimi and G. Lan. Stochastic rst- and zeroth-order methods for nonconvex stochastic programming. SIAM Journal on Optimization, 23(4):23412368, 2021. 3 9 I. J. Goodfellow, J. Shlens, and C.

7、 Szegedy. Explaining and harnessing adversarial examples. In ICLR, 2021. 1, 2 10 N. Hansen and A. Ostermeier. Completely derandomized self-adaptation in evolution strategies. Evolutionary computation, 9(2):159195, 2001. 3, 4 11 K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learning for image re

8、cognition. In CVPR, 2016. 1 12 G. B. Huang, M. Mattar, T. Berg, and E. Learned-Miller. Labeled faces in the wild: A database forstudying face recognition in unconstrained environments. In Workshop on faces inReal-LifeImages: detection, alignment, and recognition, 2020. 1, 2, 5 13 C. Igel, T. Suttorp

9、, and N. Hansen. A computational efcient covariance matrix update and a (1 1)-cma for evolution strategies. In Proceedings of the 8th annual conference on Genetic and evolutionary computation, pages 453460. ACM, 2006. 3 14 A. Ilyas, L. Engstrom, A. Athalye, and J. Lin. Black-box adversarial attacks

10、with limited queries and information. In ICML, 2018. 2, 3, 5, 6, 8 15 I. Kemelmacher-Shlizerman, S. M. Seitz, D. Miller, and E. Brossard. The megaface benchmark: 1 million faces for recognition at scale. In CVPR, 2016. 1, 2, 5 16 W. Liu, Y. Wen, Z. Yu, M. Li, B. Raj, and L. Song. Sphereface: Deep hypersphere embedding for face recognition. In CVPR, 2017. 1, 2, 5, 6 0x03 论文模型代码实现 待更 0x04 预备知识 0x04.1 协方差矩阵2, 3, 4 0x04.1.1 方差与协方差 方差 是用来度量单个随机变量的离散程度。 方差的计算公式为