ARX Block Cipher Distinguisher Based on Quantum Convolutional Neural Network

doi:10.3969/j.issn.1671-1122.2025.03.009

Abstract

Abstract:

With the development of quantum computers, quantum neural network technology continues to make new breakthroughs. Although the current quantum computing environment is still limited, exploring the potential application areas of quantum neural networks is of great significance for the development of future technologies. Quantum convolutional neural networks, which combine the advantages of quantum computing and the powerful feature extraction capabilities of neural networks, have demonstrated excellent performance in binary classification tasks. This paper proposed a quantum convolutional neural distinguisher, in which data features were encoded into the quantum circuit as a whole rather than in multiple partitions, parameterized quantum convolutional circuit was then trained. Taking SPECK-32 as an example, by used 8 qubits, the accuracy of this distinguisher which runned 5 rounds is 76.8%, surpassed the classical distinguisher under the same resource conditions, and successfully runned to 6 rounds. This paper compared quantum neural distinguishers using convolutional circuits and hardware-efficient Ansatz as training circuits, and the results indicate that the former exhibits higher efficiency. In addition, the quantum convolutional distinguisher successfully operated on reduced-round versions of Speckey, LAX32, SIMON-32 and SIMECK-32 algorithms. Finally, factors influencing the performance of the quantum convolutional neural distinguisher were analyzed.

Key words: quantum convolutional neural network, quantum computing, block cipher, distinguisher

CLC Number:

TP309

QIN Guangxue, LI Lisha. ARX Block Cipher Distinguisher Based on Quantum Convolutional Neural Network[J]. Netinfo Security, 2025, 25(3): 467-477.

Figures/Tables 17

References 31

[1]	BIAMONTE J, WITTEK P, PANCOTTI N, et al. Quantum Machine Learning[J]. Nature, 2017, 549(7671): 195-202.
[2]	RUAN Yue, CHEN Hanwu, LIU Zhihao, et al. Quantum Principal Component Analysis Algorithm[J]. Chinese Journal of Computers, 2014, 37(3): 666-676.
[3]	LI Zhaokai, LIU Xiaomei, XU Nanyang, et al. Experimental Realization of a Quantum Support Vector Machine[EB/OL]. (2015-04-08)[2024-11-19]. https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.114.140504.
[4]	DANG Yijie, JIANG Nan, HU Hao, et al. Image Classification Based on Quantum K-Nearest-Neighbor Algorithm[J]. Quantum Information Processing, 2018, 17: 1-18.
[5]	DWIVEDI K, NIGAM A. Quantum Deep Learning for Phoniatrics Biomarker Based Disease Detection Deep Learning Model for Disease Detection by Phoniatrics Biomarkers Using Quantum Computation[C]// IEEE. 2nd International Conference on Data, Engineering and Applications (IDEA). New York: IEEE, 2020: 1-8.
[6]	DONG Yumin, FU Yanying, LIU Hengrui, et al. An Improved Hybrid Quantum-Classical Convolutional Neural Network for Multi-Class Brain Tumor MRI Classification[EB/OL]. (2023-02-14)[2024-11-19]. https://pubs.aip.org/aip/jap/article-abstract/133/6/064401/2872018.
[7]	ZOUFAL C, LUCCHI A, WOERNER S. Variational Quantum Boltzmann Machines[J]. Quantum Machine Intelligence, 2021, 3: 1-15.
[8]	STAMATOPOULOS N, EGGER D J, SUN Yue, et al. Option Pricing Using Quantum Computers[EB/OL]. (2020-07-02)[2024-11-19]. https://arxiv.org/abs/1905.02666v5.
[9]	WANG Jian, ZHANG Rui, JIANG Nan. Survey on Quantum Machine Learning[J]. Journal of Software, 2024, 35(8): 3843-3877.
	王健, 张蕊, 姜楠. 量子机器学习综述[J]. 软件学报, 2024, 35(8): 3843-3877.
[10]	WIEBE N, KAPOOR A, SVORE K M. Quantum Deep Learning[EB/OL]. (2014-12-10)[2024-11-19]. https://arxiv.org/abs/1412.3489.
[11]	LLOYD S, WEEDBROOK C. Quantum Generative Adversarial Learning[EB/OL]. (2018-07-26)[2024-11-19]. https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.121.040502.
[12]	CHEN S Y C, YANG C H H, QI Jun, et al. Variational Quantum Circuits for Deep Reinforcement Learning[J]. IEEE Access, 2020, 8: 141007-141024.
[13]	CONG I, CHOI S, LUKIN M D. Quantum Convolutional Neural Networks[J]. Nature Physics, 2019, 15(12): 1273-1278. doi: 10.1038/s41567-019-0648-8
[14]	KAK S C. Quantum Neural Computing[J]. Advances in Imaging and Electron Physics. 1995, 94: 259-313.
[15]	ABBAS A, SUTTER D, ZOUFAL C, et al. The Power of Quantum Neural Networks[J]. Nature Computational Science, 2021, 1(6): 403-409. doi: 10.1038/s43588-021-00084-1 pmid: 38217237
[16]	KIM H, JANG K, LIM S, et al. Quantum Neural Network Based Distinguisher on SPECK-32/64[EB/OL]. (2023-06-18)[2024-11-19]. https://www.mdpi.com/1424-8220/23/12/5683.
[17]	HUR T, KIM L, PARK D K. Quantum Convolutional Neural Network for Classical Data Classification[EB/OL]. (2022-02-11)[2024-11-19]. https://arxiv.org/abs/2108.00661.
[18]	GOHR A. Improving Attacks on Round-Reduced SPECK32/64 Using Deep Learning[EB/OL]. (2019-08-15)[2024-11-19]. https://ia.cr/2019/037.
[19]	CHEN Yi, YU Hongbo. A New Neural Distinguisher Model Considering Derived Features from Multiple Ciphertext Pairs[EB/OL]. (2021-03-09)[2024-11-19]. https://ia.cr/2021/310.
[20]	BENAMIRA A, GERAULT D, PEYRIN T, et al. A Deeper Look at Machine Learning-Based Cryptanalysis[EB/OL]. (2021-03-22)[2024-12-19]. https://ia.cr/2021/287.
[21]	BAKSI A, BREIER J, CHEN Yi, et al. Machine Learning Assisted Differential Distinguishers for Lightweight Ciphers (Extended Version)[EB/OL]. (2020-12-02)[2024-11-19]. https://ia.cr/2020/571.
[22]	HOU Zezhou, REN Jiongjiong, CHEN Shaozhen. Cryptanalysis of Round-Reduced SIMON32 Based on Deep Learning[EB/OL]. (2021-03-18)[2024-11-19]. https://ia.cr/2021/362.
[23]	YADAV T, KUMAR M. Differential-ML Distinguisher: Machine Learning Based Generic Extension for Differential Cryptanalysis[EB/OL]. (2020-10-29)[2024-11-19]. https://ia.cr/2020/913.
[24]	LU Jinyu, LIU Guoqiang, SUN Bing, et al. Improved (Related-Key) Differential-Based Neural Distinguishers for SIMON and Simeck Block Ciphers[J]. The Computer Journal, 2024, 67(2): 537-547.
[25]	RAJAN R, ROY R K, SEN D, et al. Deep Learning-Based Differential Distinguisher for Lightweight Cipher GIFT-COFB[C]// IPSCTM. Proceedings of MISS 2021. Heidelberg: Springer, 2022: 397-406.
[26]	YANG Xiaoxue, CHEN Jie, HAN Lidong. Application of Deep Learning in Differential Cryptanalysis of ARX Block Ciphers[J]. Journal of Cryptologic Research, 2022, 9(5): 923-935.
	杨小雪, 陈杰, 韩立东. 深度学习在 ARX 分组密码差分分析的应用[J]. 密码学报, 2022, 9(5): 923-935.
[27]	HAVLÍČEK V, CÓRCOLES A D, TEMME K, et al. Supervised Learning with Quantum-Enhanced Feature Spaces[J]. Nature, 2019, 567(7747): 209-212.
[28]	BIRYUKOV A, VELICHKOV V, LE C Y. Automatic Search for the Best Trails in ARX: Application to Block Cipher Speck[EB/OL]. (2016-04-25)[2024-11-19]. https://ia.cr/2016/409.
[29]	DINU D, PERRIN L, UDOVENKO A, et al. Design Strategies for ARX with Provable Bounds: Sparx and LAX (Full Version)[EB/OL]. (2021-05-31)[2024-11-19]. https://ia.cr/2016/984.
[30]	BEAULIEU R, SHORS D, SMITH J, et al. The SIMON and SPECK Lightweight Block Ciphers[C]// IEEE. 52nd ACM/EDAC/IEEE Design Automation Conference (DAC). New York: IEEE, 2015: 1-6.
[31]	YANG Gangqiang, ZHU Bo, SUDER V, et al. The Simeck Family of Lightweight Block Ciphers[EB/OL]. (2015-06-30)[2024-11-19]. https://ia.cr/2015/612.

方法	量子比特数 /个	编码方法	参数数量 /个	数据大小/个（训练，测试）	周期数 /次	批大小 /个	训练准确率	测试准确率
文献[16]	16	基态编码	385	2^13.8727, 2^9.9660	10	32	52%	53%
本文	8	IQP编码变体	213	2^13.8727, 2^9.9660	10	32	75.9%	76.8%
Constrained C	—	—	4353	同本文	同本文	同本文	79%	76%
文献[18] （C1）	—	—	100897	2^23.2534, 2^19.9316	200	5000	93%	93%
文献[19] （C2）	—	—	自适应调整	2^23.2534, 2^19.9316	10	500	99% (最大)	99% (最大)

算法	轮数 /轮	模式	学习率	训练数据量/个	周期数/次	批大小/个	测试准确率
SPECK-32	5	Q	0.1%	2^13.8727	10	32	76.8%
SPECK-32	5	C	—	2^23.2534	10	500	99%(最大)^[19]
SPECK-32	6	Q	1%	2^13.8727	10	64	58.5%
SPECK-32	6	C	—	2^23.2534	10	500	98%(最大)^[19]
Speckey	5	Q	0.1%	2^13.8727	10	32	77.2%
Speckey	5	C	—	2^23.2534	100	5000	97.4%(最佳)^[26]
Speckey	6	Q	1%	2^13.8727	10	64	59.8%
Speckey	6	C	—	2^23.2534	100	5000	87.8%(最佳)^[26]
LAX32	3	Q	1%	2^13.8727	10	64	58.3%
LAX32	3	C	—	2^23.2534	100	5000	90.2%(最佳)^[26]

算法	轮数/轮	模式	学习率	训练数据量/个	周期数/次	批大小/个	测试准确率
SIMON-32	5	Q	1%	2^13.8727	10	64	61.70%
SIMON-32	9	C	—	2^24.2534	120	3000	91.76%^[24]
SIMECK-32	5	Q	1%	2^13.8727	10	64	85.10%
SIMECK-32	9	C	—	2^24.2534	120	3000	99.55%^[24]

方法	量子比特数 /个	训练电路量子门数/个	参数数量/个	线路层数 /层	测试准确率
HEA-QNN	8	398	216	84	A-CX (71.50%)*	A-CY (53.30%)	A-CZ (63.50%)
QCNN	8	156	213	42	76.80%

电路	参数电路量子门数/个	参数数量/个	训练准确率	测试准确率
卷积电路1	63	38	50.10%	46.50%
卷积电路2	89	51	65.00%	61.90%
卷积电路3	102	64	58.10%	60.60%
卷积电路4	102	90	65.20%	64.70%
卷积电路5	102	90	65.10%	66.50%
卷积电路6	128	90	65.40%	64.90%
卷积电路7	154	142	64.80%	66.80%
卷积电路8	154	142	64.60%	67.50%
图5卷积电路	156	213	75.90%	76.80%