基于量子局部内在维度的对抗样本检测算法

doi:10.11772/j.issn.1001-9081.2023020172

《计算机应用》唯一官方网站 ›› 2024, Vol. 44 ›› Issue (2): 490-495.DOI: 10.11772/j.issn.1001-9081.2023020172

• 网络空间安全 • 上一篇

基于量子局部内在维度的对抗样本检测算法

张瑜¹^,², 昌燕¹^,²(), 张仕斌¹^,²

^1.成都信息工程大学网络空间安全学院, 成都 610225
^2.先进密码技术与系统安全四川省重点实验室(成都信息工程大学), 成都 610225

收稿日期:2023-02-23 修回日期:2023-04-23 接受日期:2023-05-06 发布日期:2023-08-14 出版日期:2024-02-10
通讯作者: 昌燕
作者简介:张瑜（1998—），女，重庆丰都人，硕士研究生，主要研究方向：量子计算、量子机器学习隐私保护
张仕斌（1971—），男，重庆丰都人，教授，博士，CCF会员，主要研究方向：量子计算、信息安全。
基金资助:
国家自然科学基金资助项目(62272068);成都市重点研发支撑计划项目(2021?YF09?00114?GX)

Adversarial example detection algorithm based on quantum local intrinsic dimensionality

Yu ZHANG¹^,², Yan CHANG¹^,²(), Shibin ZHANG¹^,²

^1.School of Cybersecurity，Chengdu University of Information Technology，Chengdu Sichuan 610225，China
^2.Sichuan Provincial Key Laboratory of Advanced Cryptography and System Security （Chengdu University of Information Technology），Chengdu Sichuan 610225，China

Received:2023-02-23 Revised:2023-04-23 Accepted:2023-05-06 Online:2023-08-14 Published:2024-02-10
Contact: Yan CHANG
About author:ZHANG Yu， born in 1998， M. S. candidate. Her research interests include quantum computing， privacy protection in quantum machine learning.
ZHANG Shibin， born in 1971， Ph. D.， professor. His research interests include quantum computing， information security.
Supported by:
National Natural Science Foundation of China(62272068);Key Research and Development Support Plan of Chengdu(2021-YF09-00114-GX)

摘要/Abstract

摘要：

为解决基于局部内在维度（LID）的对抗样本检测算法高时间复杂度问题，结合量子计算优势，提出一种基于量子LID的对抗样本检测算法。首先，使用SWAP-Test量子算法一次性计算待测样本与所有样本间的相似度，避免了经典算法中的冗余计算；然后，结合量子相位估计（QPE）算法和量子Grover搜索算法计算待测样本的局部内在维度；最后，以LID作为二分类检测器的评判依据，检测区分出对抗样本。分别使用IRIS、MNIST、股票时序数据集测试和验证所提算法，仿真实验结果表明，均能通过计算出的LID值突出对抗样本与正常样本之间的差异性，并能作为检测依据区分样本属性。理论研究证明，所提算法时间复杂度与Grover算子迭代次数及邻近样本数和训练样本数的平方根的积同一数量级，明显优于基于LID的对抗样本检测算法，实现了指数级加速。

关键词: 局部内在维度, 对抗样本, 二分类检测, 量子计算, 股价预测

Abstract:

In order to solve the high time complexity problem of the adversarial example detection algorithm based on Local Intrinsic Dimensionality （LID）， combined with the advantages of quantum computing， an adversarial example detection algorithm based on quantum LID was proposed. First， the SWAP-Test quantum algorithm was used to calculate the similarity between the measured example and all examples in one time， avoiding the redundant calculation in the classical algorithm. Then Quantum Phase Estimation （QPE） algorithm and quantum Grover search algorithm were combined to calculate the local intrinsic dimension of the measured example. Finally， LID was used as the evaluation basis of the binary detector to detect and distinguish the adversarial examples. The detection algorithm was tested and verified on IRIS， MNIST， and stock time series datasets. The simulation experimental results show that the calculated LID values can highlight the difference between adversarial examples and normal examples， and can be used as a detection basis to differentiate example attributes. Theoretical research proves that the time complexity of the proposed detection algorithm is the same order of magnitude as the product of the number of iterations of Grover operator and the square root of the number of adjacent examples and the number of training examples， which is obviously better than that of the adversarial example detection algorithm based on LID and achieves exponential acceleration.

Key words: Local Intrinsic Dimensionality (LID), adversarial example, binary detection, quantum computing, stock prediction

中图分类号:

TP391

张瑜, 昌燕, 张仕斌. 基于量子局部内在维度的对抗样本检测算法[J]. 计算机应用, 2024, 44(2): 490-495.

Yu ZHANG, Yan CHANG, Shibin ZHANG. Adversarial example detection algorithm based on quantum local intrinsic dimensionality[J]. Journal of Computer Applications, 2024, 44(2): 490-495.

图/表 7

图1 SWAP-Test量子线路

Fig. 1 SWAP-Test quantum circuit

图2 SWAP门线路

Fig. 2 SWAP-gate circuit

图3 多分类DNN模型

Fig. 3 Multi-classification DNN model

图4 基于量子LID的对抗样本检测算法程序实现

Fig. 4 Implementation of adversarial example detection algorithm based on quantum LID

图5 使用经典LID和量子LID算法计算IRIS数据样本LID值的实验结果对比

Fig. 5 Experimental result comparison of calculating LID values of IRIS data examples by classical LID and quantum LID algorithms

图6 使用量子LID算法计算MNIST数据样本LID值的实验结果

Fig. 6 Experimental results of calculating LID values of MNIST data examples by quantum LID algorithm

表1 使用量子LID算法计算股票数据样本LID值的实验结果

Tab. 1 Experimental results of calculating LID values of stock data examples by quantum LID algorithm

样本实例	LID估计值
样本实例	正常样本	对抗样本
1	298.76	707.95
2	163.93	239.64
3	472.52	599.19

参考文献 20

1	郑远攀，李广阳，李晔. 深度学习在图像识别中的应用研究综述［J］. 计算机工程与应用， 2019， 55（12）： 20-36. 10.3778/j.issn.1002-8331.1903-0031
	ZHENG Y P， LI G Y， LI Y. Survey of application of deep learning in image recognition［J］. Computer Engineering and Applications， 2019， 55（12）： 20-36. 10.3778/j.issn.1002-8331.1903-0031
2	PURWINS H， LI B， VIRTANEN T. Deep learning for audio signal processing［J］. IEEE Journal of Selected Topics in Signal Processing， 2019， 13（2）： 206-219. 10.1109/jstsp.2019.2908700
3	DAI H， KHALIL E B， ZHANG Y， et al. Learning combinatorial optimization algorithms over graphs［C］// Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook， NY： Curran Associates Inc.， 2017： 6351-6361.
4	JIANG W. Applications of deep learning in stock market prediction： recent progress［J］. Expert Systems with Applications， 2021， 184： No.115537. 10.1016/j.eswa.2021.115537
5	张晓旭，高振涛，吴磊，等. 基于混合量子-经典神经网络模型的股价预测［J］. 电子科技大学学报， 2022， 51（1）：16-23.
	ZHANG X X， GAO Z T， WU L， et al. Stock price prediction based on a hybrid quantum-classical neural network model［J］. Journal of University of Electronic Science and Technology of China， 2022， 51（1）：16-23.
6	GOODFELLOW I J， SHLENS J， SZEGEDY C. Explaining and harnessing adversarial examples［EB/OL］. （2015-03-20）［2022-07-23］..
7	MOOSAVI-DEZFOOLI S M， FAWZI A， FROSSARD P. DeepFool： a simple and accurate method to fool deep neural networks［C］// Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2016： 2574-2582. 10.1109/cvpr.2016.282
8	PAPERNOT N， McDANIEL P， JHA S， et al. The limitations of deep learning in adversarial settings［C］// Proceedings of the 2016 IEEE European Symposium on Security and Privacy. Piscataway： IEEE， 2016： 372-387. 10.1109/eurosp.2016.36
9	CARLINI N， WAGNER D. Towards evaluating the robustness of neural networks［C］// Proceedings of the 2017 IEEE Symposium on Security and Privacy. Piscataway： IEEE， 2017： 39-57. 10.1109/sp.2017.49
10	MIYATO T， DAI A M， GOODFELLOW I. Adversarial training methods for semi-supervised text classification［EB/OL］. （2021-11-16）［2022-07-23］..
11	ATHALYE A， CARLINI N， WAGNER D. Obfuscated gradients give a false sense of security： circumventing defenses to adversarial examples［C］// Proceedings of the 35th International Conference on Machine Learning. New York： JMLR.org， 2018： 274-283. 10.48550/arXiv.1802.00420
12	LIAO F， LIANG M， DONG Y， et al. Defense against adversarial attacks using high-level representation guided denoiser［C］// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： Piscataway： IEEE， 2018： 1778-1787. 10.1109/cvpr.2018.00191
13	PAPERNOT N， McDANIEL P， WU X， et al. Distillation as a defense to adversarial perturbations against deep neural networks［C］// Proceedings of the 2016 IEEE Symposium on Security and Privacy. Piscataway： IEEE， 2016： 582-597. 10.1109/sp.2016.41
14	XU W， EVANS D， QI Y. Feature squeezing： detecting adversarial examples in deep neural networks［C］// Proceedings of the Network and Distributed System Security Symposium 2018. Reston， VA： Internet Society， 2018： No.23198. 10.14722/ndss.2018.23198
15	FEINMAN R， CURTIN R R， SHINTRE S， et al. Detecting adversarial samples from artifacts［EB/OL］. （2017-11-15）［2022-07-23］..
16	PANG T， DU C， ZHU J. Robust deep learning via reverse cross-entropy training and thresholding test［EB/OL］. （2017-06-02）［2022-07-23］..
17	MA X， LI B， WANG Y， et al. Characterizing adversarial subspaces using local intrinsic dimensionality［EB/OL］. （2018-03-14）［2022-07-23］..
18	阮越，陈汉武，刘志昊，等. 量子主成分分析算法［J］. 计算机学报， 2014， 37（3）： 666-676.
	RUAN Y， CHEN H W， LIU Z H， et al. Quantum principal component analysis algorithm［J］. Chinese Journal of Computers， 2014， 37（3）： 666-676.
19	CONG I， CHOI S， LUKIN M D. Quantum convolutional neural networks［J］. Nature Physics， 2019， 15（12）： 1273-1278. 10.1038/s41567-019-0648-8
20	AMSALEG L， BAILEY J， BARBE D， et al. The vulnerability of learning to adversarial perturbation increases with intrinsic dimensionality［C］// Proceedings of the 2017 IEEE Workshop on Information Forensics and Security. Piscataway： IEEE， 2017： 1-6. 10.1109/wifs.2017.8267651

[1]	陈彤, 位纪伟, 何仕远, 宋井宽, 杨阳. 基于自适应攻击强度的对抗训练方法[J]. 《计算机应用》唯一官方网站, 2024, 44(1): 94-100.
[2]	张济慈, 范纯龙, 李彩龙, 郑学东. 基于几何关系的跨模型通用扰动生成方法[J]. 《计算机应用》唯一官方网站, 2023, 43(11): 3428-3435.
[3]	李宇航, 杨玉丽, 马垚, 于丹, 陈永乐. 基于BERT模型的文本对抗样本生成方法[J]. 《计算机应用》唯一官方网站, 2023, 43(10): 3093-3098.
[4]	魏佳璇, 杜世康, 于志轩, 张瑞生. 图像分类中的白盒对抗攻击技术综述[J]. 《计算机应用》唯一官方网站, 2022, 42(9): 2732-2741.
[5]	杨博, 张恒巍, 李哲铭, 徐开勇. 基于图像翻转变换的对抗样本生成方法[J]. 《计算机应用》唯一官方网站, 2022, 42(8): 2319-2325.
[6]	王曙燕, 侯则昱, 孙家泽. 面向深度学习的对抗样本差异性检测方法[J]. 计算机应用, 2021, 41(7): 1849-1856.
[7]	吴立人, 刘政浩, 张浩, 岑悦亮, 周维. 聚焦图像对抗攻击算法PS-MIFGSM[J]. 计算机应用, 2020, 40(5): 1348-1353.
[8]	赵学武刘广亮程新党冀俊忠. 基于拓扑序列和量子遗传算法的贝叶斯网结构学习[J]. 计算机应用, 2013, 33(06): 1595-1603.
[9]	易正俊何荣花侯坤. 量子位Bloch坐标的量子人工蜂群优化算法[J]. 计算机应用, 2012, 32(07): 1935-1938.
[10]	付国瑜黄贤英. 基于Web页面平均质量的Web搜索模型和优化算法[J]. 计算机应用, 2009, 29(4): 1114-1116.
[11]	武妍包建军 . 一种新的求解TSP的混合量子进化算法[J]. 计算机应用, 2006, 26(10): 2433-2436.

基于量子局部内在维度的对抗样本检测算法

Adversarial example detection algorithm based on quantum local intrinsic dimensionality

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 7

参考文献 20

相关文章 11

编辑推荐

Metrics