南京大学学报(自然科学版) ›› 2021, Vol. 57 ›› Issue (5): 793–800.doi: 10.13232/j.cnki.jnju.2021.05.009
• • 上一篇
基于生成对抗网络的音频目标分类对抗
张强1, 杨吉斌2(
), 张雄伟2, 曹铁勇2, 梅鹏程1
- 1.陆军工程大学研究生院，南京，210007
2.陆军工程大学指挥控制工程学院，南京，210007
Generating adversarial examples in audio object classification using generative adversarial network
Qiang Zhang1, Jibin Yang2(), Xiongwei Zhang2, Tieyong Cao2, Pengcheng Mei1
- 1.Graduate School，Army Engineering University，Nanjing，210007，China
2.School of Command and Control Engineering，Army Engineering University，Nanjing，210007，China
摘要：
音频对抗样本可以用于提高音频目标分类系统的可靠性，然而目前音频对抗样本的感知质量较低，生成质量不能令人满意.为提升音频对抗样本的质量，首次采用生成对抗网络（Generative Adversarial Network，GAN）实现音频目标分类的对抗样本生成.提出用于音频目标分类对抗样本生成的通用GAN框架，将待攻击的分类模型引入GAN.在此基础上，提出基于GAN的分段扰动/整体攻击（GAN?based Segmented?perturbation Overall?attack，SOGAN）方法.SOGAN通过对抗训练，学习短时分段音频数据上的有效扰动，按照与原始音频的对应关系生成整体扰动，并形成时长可变的对抗样本.该方法可以缩小音频对抗样本的搜索空间，降低对抗样本生成的复杂度.在UrbanSound8k，ESC50等音频目标分类数据集上的实验表明，和已有音频目标对抗样本设计方法相比，所提方法生成的对抗样本可感知性更低，对典型音频目标分类系统具有较高的攻击成功率和攻击效率.
中图分类号:
- TP391
| 1 | Khamparia A，Gupta D，Nguyen N G，et al. Sound classification using convolutional neural network and tensor deep stacking network. IEEE Access，2019(7)：7717-7727. |
| 2 | Hannun A，Case C，Casper J，et al. Deep speech：Scaling up end?to?end speech recognition. 2014，arXiv:. |
| 3 | Hoshen Y，Weiss R J，Wilson K W. Speech acoustic modeling from raw multichannel waveforms∥2015 IEEE International Conference on Acoustics，Speech and Signal Processing. South Brisbane，Australia：IEEE，2015：4624-4628. |
| 4 | Van Den Oord A，Dieleman S，Zen H，et al. Wavenet：A generative model for raw audio. 2016，arXiv:. |
| 5 | Sainath T N，Weiss R J，Senior A，et al. Learning the speech front?end with raw waveform CLDNNs∥The 16th Annual Conference of the International Speech Communication Association. Dresden，Germany：IEEE，2015：1-5. |
| 6 | Szegedy C，Zaremba W，Sutskever I，et al. Intriguing properties of neural networks. 2013,arXiv：1312. 6199v4. |
| 7 | Goodfellow I J，Shlens J，Szegedy C. Explaining and harnessing adversarial examples. 2014,arXiv：1412. 6572v2. |
| 8 | Kurakin A，Goodfellow I J，Bengio S. Adversarial examples in the physical world. 2016,arXiv：1607. 02533v1. |
| 9 | Carlini N，Wagner D. Towards evaluating the robustness of neural networks∥2017 IEEE Symposium on Security and Privacy. San Jose，CA，USA：IEEE，2017：39-57. |
| 10 | Papernot N，McDaniel P，Jha S，et al. The limitations of deep learning in adversarial settings∥2016 IEEE European Symposium on Security and Privacy. Saarbruecken，Germany：IEEE，2016：372-387. |
| 11 | Moosavi?Dezfooli S M，Fawzi A，Frossard P. Deepfool：A simple and accurate method to fool deep neural networks∥2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas，NV，USA：IEEE，2016：2574-2582. |
| 12 | Hu W W，Tan Y. Generating adversarial malware examples for black?box attacks based on GAN. 2017,arXiv：. |
| 13 | Moosavi?Dezfooli S M，Fawzi A，Fawzi O，et al. Universal adversarial perturbations∥2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu，HI，USA：IEEE，2017：86-94. |
| 14 | Xiao C W，Li B，Zhu J Y，et al. Generating adversarial examples with adversarial networks∥Proceedings of the 27th International Joint Conference on Artificial Intelligence. Stockholm,Sweden：AAAI Press，2018：3905-3911. |
| 15 | Carlini N，Wagner D. Audio adversarial examples：Targeted attacks on speech?to?text∥2018 IEEE Security and Privacy Workshops. San Francisco，CA，USA：IEEE，2018：1-7. |
| 16 | Alzantot M，Balaji B，Srivastava M. Did you hear that?Adversarial examples against automatic speech recognition. 2018,arXiv：. |
| 17 | Du T Y，Ji S L，Li J F，et al. SirenAttack：Generating adversarial audio for end?to?end acoustic systems∥Proceedings of the 15th ACM Asia Conference on Computer and Communications Security. New York，NY,USA：ACM，2020：357-369. |
| 18 | Abdoli S，Hafemann L G，Rony J，et al. Universal adversarial audio perturbations. 2019,arXiv:1908. 03173. |
| 19 | Wang D H，Dong L，Wang R D，et al. Targeted speech adversarial example generation with generative adversarial network. IEEE Access，2020(8)：124503-124513. |
| 20 | Goodfellow I J，Pouget?Abadie J，Mirza M，et al. Generative adversarial networks. 2014,arXiv:1406. 2661. |
| 21 | Salamon J，Jacoby C，Bello J P. A dataset and taxonomy for urban sound research∥Proceedings of the 22nd ACM International Conference on Multimedia. New York，NY，USA：ACM，2014：1041-1044. |
| 22 | Piczak K J. ESC：Dataset for environmental sound classification∥Proceedings of the 23rd ACM International Conference on Multimedia. New York，NY，USA：ACM，2015：1015-1018. |
| 23 | Johnson J，Alahi A，Li F F. Perceptual losses for real?time style transfer and super?resolution∥European Conference on Computer Vision. Springer Berlin Heidelberg，2016：694-711. |
| 24 | Tokozume Y，Harada T. Learning environmental sounds with end?to?end convolutional neural network∥2017 IEEE International Conference on Acoustics，Speech and Signal Processing. New Orleans，LA，USA：IEEE，2017：2721-2725. |
| 25 | Paszke A，Gross S，Massa F，et al. PyTorch：An imperative style，high?performance deep learning library. 2019，arXiv：. |
| 26 | Tokozume Y，Ushiku Y，Harada T. Learning from between?class examples for deep sound recognition. 2017，arXiv：. |
| [1] | 房笑宇, 曹陈涵, 夏彬. 基于注意力机制的大规模系统日志异常检测方法[J]. 南京大学学报(自然科学版), 2021, 57(5): 785-792. |
| [2] | 戴臣超, 王洪元, 曹亮, 殷雨昌, 张继. 一种多目标跨摄像头跟踪技术研究与实现[J]. 南京大学学报(自然科学版), 2021, 57(2): 227-236. |
|
||