基于流形学习能量数据预处理的模板攻击优化方法

袁庆军; 王安; 王永娟; 王涛

doi:10.11999/JEIT190598

引用本文: 袁庆军, 王安, 王永娟, 王涛. 基于流形学习能量数据预处理的模板攻击优化方法[J]. 电子与信息学报, doi: 10.11999/JEIT190598 shu

Citation: Qingjun YUAN, An WANG, Yongjuan WANG, Tao WANG. An Improved Template Analysis Method Based on Power Traces Preprocessing with Manifold Learning[J]. Journal of Electronics and Information Technology, doi: 10.11999/JEIT190598 shu

基于流形学习能量数据预处理的模板攻击优化方法

1.
战略支援部队信息工程大学郑州 450001
2.
河南省网络密码技术重点实验室郑州 450001
3.
北京理工大学计算机学院北京 100081

作者简介: 袁庆军: 男，1993年生，助教教授，研究方向为机器学习侧信道分析;
王安: 男，1983年生，副教授，研究方向为侧信道分析与防护技术;
王永娟: 女，1982年生，研究员，研究方向为侧信道分析与密码系统安全;
王涛: 男，1995年生，硕士生，研究方向为机器学习侧信道分析

通讯作者: 王永娟,pinkywyj@163.com

基金项目: 国家自然科学基金(61872040)，河南省网络密码技术重点实验室开放基金(LNCT2019-S02)，“十三五”国家密码发展基金(MMJJ20170201)

摘要: 能量数据作为模板攻击过程中的关键对象，具有维度高、有效维度少、不对齐的特点，在进行有效的预处理之前，模板攻击难以奏效。针对能量数据的特性，该文提出一种基于流形学习思想进行整体对齐的方法，以保留能量数据的变化特征，随后通过线性投影的方法降低数据的维度。使用该方法在Panda 2018 challenge1标准数据集进行了验证，实验结果表明，该方法的特征提取效果优于传统的PCA和LDA方法，能大幅度提高模板攻击的成功率。最后采用模板攻击恢复密钥，仅使用两条能量迹密钥恢复成功率即可达到80%以上。

关键词:

English

An Improved Template Analysis Method Based on Power Traces Preprocessing with Manifold Learning

1.
PLA Strategic Support Force Information Engineering University , Zhengzhou 450001, China
2.
Henan Key Laboratory of Network Cryptography Technology, Zhengzhou 450001, China
3.
School of Computer Science, Beijing Institute of Technology, Beijing 100081, China

Corresponding author: Yongjuan WANG,pinkywyj@163.com

Received Date: 2019-08-07
Accepted Date: 2019-10-31
Available Online: 2019-01-01
Fund Project: The National Natural Science Foundation of China (61872040), The Fund of Henan Key Laboratory of Network Cryptography Technology (LNCT2019-S02), The National Cryptographic Development Fund of the 13th Five-Year Plan (MMJJ20170201)

Abstract: As the key object in the process of template analysis, power traces have the characteristics of high dimension, less effective dimension and unaligned. Before effective preprocessing, template attack is difficult to work. Based on the characteristics of energy data, a global alignment method based on manifold learning is proposed to preserve the changing characteristics of power traces, and then the dimensionality of data is reduced by linear projection. The method is validated in Panda 2018 challenge1 standard datasets respectively. The experimental results show that the feature extraction effect of this method is superior than that of traditional PCA and LDA methods. Finally, the method of template analysis is used to recover the key, and the recovery success rates can reach 80% with only two traces.

Key words:

1. [1]
  KOCHER P, JAFFE J, and JUN B. Differential power analysis[C]. The 13th Annual International Cryptology Conference, Santa Barbara, USA, 1999: 388–397. doi: 10.1007/3-540-48405-1_25.
2. [2]
  ERNST D and MARTIN S. The common criteria for information technology security evaluation: Implications for China’s policy on information security standards[R]. East-West Center Working Papers, No. 108, 2010. doi: 10.2139/ssrn.2770146.
3. [3]
  VAN TILBORG H C A AND JAJODIA S. Encyclopedia of Cryptography and Security[M]. Boston: Springer, 2011: 468–471. doi: 10.1007/978-1-4419-5906-5.
4. [4]
  CHARI S, RAO J R, and ROHATGI P. Template attacks[C]. The 4th International Workshop on Cryptographic Hardware and Embedded Systems, Redwood Shores, USA, 2002: 13–28. doi: 10.1007/3-540-36400-5_3.
5. [5]
  BRIER E, CLAVIER C, and OLIVIER F. Correlation power analysis with a leakage model[C]. The 6th International Workshop on Cryptographic Hardware and Embedded Systems, Cambridge, USA, 2004: 16–29. doi: 10.1007/978-3-540-28632-5_2.
6. [6]
  BOGDANOV A. Improved side-channel collision attacks on AES[C]. The 14th International Workshop on Selected Areas in Cryptography, Ottawa, Canada, 2007: 84–95. doi: 10.1007/978-3-540-77360-3_6.
7. [7]
  RIVAIN M, PROUFF E, and DOGET J. Higher-order masking and shuffling for software implementations of block ciphers[C]. The 11th International Workshop on Cryptographic Hardware and Embedded Systems, Lausanne, Switzerland, 2009: 171–188. doi: 10.1007/978-3-642-04138-9_13.
8. [8]
  CORON J S and KIZHVATOV I. Analysis and improvement of the random delay countermeasure of CHES 2009[C]. The 12th International Workshop on Cryptographic Hardware and Embedded Systems, Santa Barbara, USA, 2010: 95–109. doi: 10.1007/978-3-642-15031-9_7.
9. [9]
  黄海, 冯新新, 刘红雨, 等. 基于随机加法链的高级加密标准抗侧信道攻击对策[J]. 电子与信息学报, 2019, 41(2): 348–354. doi: 10.11999/JEIT171211
  HUANG Hai, FENG Xinxin, LIU Hongyu, et al. Random addition-chain based countermeasure against side-channel attack for advanced encryption standard[J]. Journal of Electronics &Information Technology, 2019, 41(2): 348–354. doi: 10.11999/JEIT171211
10. [10]
  LERMAN L, BONTEMPI G, and MARKOWITCH O. Power analysis attack: An approach based on machine learning[J]. International Journal of Applied Cryptography, 2014, 3(2): 97–115. doi: 10.1504/IJACT.2014.062722
11. [11]
  ARCHAMBEAU C, PEETERS E, STANDAERT F X, et al. Template attacks in principal subspaces[C]. The 8th International Workshop on Cryptographic Hardware and Embedded Systems, Yokohama, Japan, 2006: 1–14. doi: 10.1007/11894063_1.
12. [12]
  STANDAERT F X and ARCHAMBEAU C. Using subspace-based template attacks to compare and combine power and electromagnetic information leakages[C]. The 10th International Workshop on Cryptographic Hardware and Embedded Systems, Washington, USA, 2008: 411–425. doi: 10.1007/978-3-540-85053-3_26.
13. [13]
  HETTWER B, GEHRER S, and GÜNEYSU T. Applications of machine learning techniques in side-channel attacks: A survey[J]. Journal of Cryptographic Engineering, 2019. doi: 10.1007/s13389-019-00212-8
14. [14]
  王燚, 吴震, 蔺冰. 对加掩加密算法的盲掩码模板攻击[J]. 通信学报, 2019, 40(1): 1–14. doi: 10.11959/j.issn.1000-436x.2019007
  WANG Yi, WU Zhen, and LIN Bing. Blind mask template attacks on masked cryptographic algorithm[J]. Journal on Communications, 2019, 40(1): 1–14. doi: 10.11959/j.issn.1000-436x.2019007
15. [15]
  CAGLI E, DUMAS C, and PROUFF E. Convolutional neural networks with data augmentation against jitter-based countermeasures: Profiling attacks without pre-processing[C]. The 19th International Conference on Cryptographic Hardware and Embedded Systems, Taipei, China, 2017: 45–68. doi: 10.1007/978-3-319-66787-4_3.
16. [16]
  ZHOU Yuanyuan and STANDAERT F X. Deep learning mitigates but does not annihilate the need of aligned traces and a generalized ResNet model for side-channel attacks[J]. Journal of Cryptographic Engineering, 2019. doi: 10.1007/s13389-019-00209-3
17. [17]
  WANG Z. The data of PANDA challeng1[EB/OL]. https://github.com/kistoday/Panda2018/tree/master/challeng1, 2019.
18. [18]
  CRIMINISI A, SHOTTON J, and KONUKOGLU E. Decision forests: A unified framework for classification, regression, density estimation, manifold learning and semi-supervised learning[J]. Foundations and Trends® in Computer Graphics and Vision, 2012, 7(2/3): 81–227. doi: 10.1561/0600000035
19. [19]
  HOMMA N, NAGASHIMA S, IMAI Y, et al. High-resolution side-channel attack using phase-based waveform matching[C]. The 8th International Workshop on Cryptographic Hardware and Embedded Systems - CHES 2006, Yokohama, Japan, 2006: 187–200. doi: 10.1007/11894063_15.
20. [20]
  GUILLEY S, KHALFALLAH K, LOMNE V, et al. Formal framework for the evaluation of waveform resynchronization algorithms[C]. The 5th IFIP WG 11.2 International Workshop on Information Security Theory and Practice. Security and Privacy of Mobile Devices in Wireless Communication, Heraklion, Greece, 2011: 100–115. doi: 10.1007/978-3-642-21040-2_7.
21. [21]
  MATEOS E and GEBOTYS C H. A new correlation frequency analysis of the side channel[C]. The 5th Workshop on Embedded Systems Security, Scottsdale, USA, 2010: 4. doi: 10.1145/1873548.1873552.
22. [22]
  GIERLICHS B, LEMKE-RUST K, and PAAR C. Templates vs. stochastic methods: A performance analysis for side channel cryptanalysis[C]. The 8th International Workshop on Cryptographic Hardware and Embedded Systems, Yokohama, Japan, 2006: 15–29. doi: 10.1007/11894063_2.
23. [23]
  ZHANG Hailong and ZHOU Yongbin. Template attack vs. stochastic model: An empirical study on the performances of profiling attacks in real scenarios[J]. Microprocessors and Microsystems, 2019, 66: 43–54. doi: 10.1016/j.micpro.2019.02.010
1. [1]
  王永娟, 王涛, 袁庆军, 高杨, 王相宾. 密码算法旁路立方攻击改进与应用. 电子与信息学报,
2. [2]
  高东, 梁子林. 基于能量效率的双层非正交多址系统资源优化算法. 电子与信息学报,
3. [3]
  王年, 胡旭阳, 朱凡, 唐俊. 基于视图感知的单视图三维重建算法. 电子与信息学报,
4. [4]
  刘新, 阎焜, 杨光耀, 叶盛波, 张群英, 方广有. UWB-MIMO穿墙雷达三维成像与运动补偿算法研究. 电子与信息学报,
5. [5]
  高巍, 蒋刚毅, 郁梅, 骆挺. 基于熵编码的立体视频加密与信息隐藏算法. 电子与信息学报,
6. [6]
  夏晓峰, 向宏, 肖震宇, 蔡挺. 基于国产密码算法的数控网络的双层安全防护模型研究及安全评估. 电子与信息学报,
7. [7]
  毛秀海, 李凡, 左小磊. DNA数据存储. 电子与信息学报,
8. [8]
  唐伦, 曹睿, 廖皓, 王兆堃. 基于深度强化学习的服务功能链可靠部署算法. 电子与信息学报,
9. [9]
  吕敬祥, 罗文浪. 无线传感网络量化及能量优化策略. 电子与信息学报,
10. [10]
  袁庆军, 张勋成, 高杨, 王永娟. 轻量级分组密码PUFFIN的差分故障攻击. 电子与信息学报,
11. [11]
  陈前斌, 管令进, 李子煜, 王兆堃, 杨恒, 唐伦. 基于深度强化学习的异构云无线接入网自适应无线资源分配算法. 电子与信息学报,
12. [12]
  王刚, 靳彦青, 彭华, 张光伟. Lempel-Ziv-Welch压缩数据的误码纠正. 电子与信息学报,
13. [13]
  徐瑨, 吴慧慈, 陶小峰. 5G网络空间安全对抗博弈. 电子与信息学报,
14. [14]
  江小平, 王妙羽, 丁昊, 李成华. 基于信道状态信息幅值-相位的被动式室内指纹定位. 电子与信息学报,
15. [15]
  孙子文, 叶乔. 利用震荡环频率特性提取多位可靠信息熵的物理不可克隆函数研究. 电子与信息学报,
16. [16]
  谢永, 李香, 张松松, 吴黎兵. 一种可证安全的车联网无证书聚合签名改进方案. 电子与信息学报,
17. [17]
  田俊峰, 井宣. 多方参与高效撤销组成员的共享数据审计方案. 电子与信息学报,
18. [18]
  王立辉, 闫守礼, 李清. 一种轻量级数据加密标准循环掩码实现方案. 电子与信息学报,
19. [19]
  左志斌, 常朝稳, 祝现威. 一种基于数据平面可编程的软件定义网络报文转发验证机制. 电子与信息学报,
20. [20]
  张波, 黄开枝, 林胜斌, 易鸣, 陈亚军. MIMO异构网络中一种基于人工噪声的抗主动窃听者的鲁棒安全传输方案. 电子与信息学报,

图 1 PANDA 2018 Challenge1 前3条能量迹

下载: 全尺寸图片幻灯片

图 2 PANDA 2018 Challenge1 能量迹与密钥相关系数

下载: 全尺寸图片幻灯片

图 3 PANDA 2018 Challenge1能量数据对齐后

下载: 全尺寸图片幻灯片

图 4 PANDA 2018 Challenge1 能量迹降维后

下载: 全尺寸图片幻灯片

图 5 PANDA 2018 Challenge1 能量迹PCA-20或LDA-20降维后

下载: 全尺寸图片幻灯片

表 1 向量矩阵计算算法

　输入：能量数据${T_\alpha } = {\rm{\{ } }{T_i},0 \le i \le \alpha ,i \in N\}$，对齐参数$k$。

　输出：对齐后的能量数据${T'_\alpha }$

　(1)　for j in range(α), do

　(2)　　计算与${T_j}$ 欧式距离最近的$k$条能量迹${\rm{\{ }}{T_{j1}},{T_{j2}}, ··· ,{T_{jk}}\} $;

　(3)　end

　(4)　for j in range (α), do

　(5)　　计算关系向量矩阵${ {{W} }_{{j} } } = \frac{ {\left( { {{C} }_i^{ - 1} \cdot { {{1} }_k} } \right)} }{ { {{1} }_k^{\rm T} \cdot {{C} }_i^{ - 1} \cdot { {{1} }_k} } }$，其中${ { C}_i} $为
　　　 ${\rm{\{ }}{T_{j1}},{T_{j2}}, ··· ,{T_{jk}}\} $的协方差矩阵，${{{1}}_k}$为$k$维全1向量；

　(6)　end

　(7)　计算矩阵${{M}} = ({{{\rm I}}} - {{W}}){({{I}} - {{W}})^{\rm{T}}}$;

　(8)　设$\beta = \alpha /2$从矩阵M中选择较小的$\beta $个特征值，记为${{{M}}_\beta }$，
　　　计算${T'_\alpha } = T \cdot {{{M}}_\beta }$;

　(9)　return ${T_\alpha }^\prime $。

下载: 导出CSV

表 2 PANDA 2018 Challenge1数据集预处理后方差表(汉明重量不同)(万)

方差	0	1	3	7	15	31	63	127	255
0	4.08	10.99	14.31	16.61	9.80	15.80	18.32	13.02	10.19
1	10.99	2.67	12.49	8.83	7.34	9.50	11.48	5.00	6.33
3	14.31	12.49	8.53	13.62	15.21	12.67	11.73	13.00	15.81
7	16.61	8.83	13.62	3.62	16.24	8.13	11.60	4.99	10.73
15	9.80	7.34	15.21	16.24	4.23	12.21	12.85	9.23	9.84
31	15.80	9.50	12.67	8.13	12.21	4.17	11.62	8.86	9.61
63	18.32	11.48	11.73	11.60	12.85	11.62	4.54	9.26	9.73
127	13.02	5.00	13.00	4.99	9.23	8.86	9.26	1.97	5.23
255	10.19	6.33	15.81	10.73	9.84	9.61	9.73	5.23	4.26

下载: 导出CSV

表 3 PANDA 2018 Challenge1数据集预处理后方差表(汉明重量相同)(万)

方差	7	11	13	14	19	35	67	131	224
7	3.62	11.23	23.70	12.19	13.35	13.52	11.55	14.04	9.86
11	11.23	2.60	18.80	11.73	12.07	11.85	12.43	10.97	10.21
13	23.70	18.80	31.91	23.04	27.09	22.52	23.58	56.33	19.22
14	12.19	11.73	23.04	3.89	12.54	9.52	14.47	14.96	12.70
19	13.35	12.07	27.09	12.54	4.78	13.86	15.33	17.68	11.98
35	13.52	11.85	22.52	9.52	13.86	3.15	15.07	15.10	10.67
67	11.55	12.43	23.58	14.47	15.33	15.07	4.98	17.73	9.50
131	14.04	10.97	56.33	14.96	17.68	15.10	17.73	37.04	20.31
224	9.86	10.21	19.22	12.70	11.98	10.67	9.50	20.31	3.91

下载: 导出CSV

表 4 PANDA 2018 Challenge1数据集PCA-20处理后方差表(汉明重量不同)(万)

方差	0	1	3	7	15	31	63	127	255
0	33.00	27.97	30.58	29.58	28.96	30.91	29.07	31.04	31.06
1	27.97	13.72	15.97	16.05	15.23	16.10	15.99	20.49	14.26
3	30.58	15.97	13.79	16.97	15.97	17.57	15.58	23.60	16.56
7	29.58	16.05	16.97	17.04	16.70	17.60	17.34	22.65	17.31
15	28.96	15.23	15.97	16.70	14.53	16.83	16.07	21.60	16.43
31	30.91	16.10	17.57	17.60	16.83	16.64	16.65	22.57	17.06
63	29.07	15.99	15.58	17.34	16.07	16.65	15.41	22.27	16.76
127	31.04	20.49	23.60	22.65	21.60	22.57	22.27	24.36	22.35
255	31.06	14.26	16.56	17.31	16.43	17.06	16.76	22.35	13.91

下载: 导出CSV

表 5 PANDA 2018 Challenge1数据集LDA-20处理后方差表(汉明重量不同)(万)

方差	0	1	3	7	15	31	63	127	255
0	0.95	1.21	0.93	0.99	1.07	1.09	1.08	1.12	1.13
1	1.21	1.13	1.07	1.17	1.20	1.11	1.24	1.15	1.20
3	0.93	1.07	0.65	0.90	0.99	0.93	1.00	1.05	1.01
7	0.99	1.17	0.90	0.84	0.97	1.02	1.10	1.09	1.06
15	1.07	1.20	0.99	0.97	0.92	1.08	1.17	1.16	1.11
31	1.09	1.11	0.93	1.02	1.08	0.89	1.10	1.10	1.02
63	1.08	1.24	1.00	1.10	1.17	1.10	1.07	1.18	1.15
127	1.12	1.15	1.05	1.09	1.16	1.10	1.18	0.98	1.15
255	1.13	1.20	1.01	1.06	1.11	1.02	1.15	1.15	0.97

下载: 导出CSV

点击查看大图

图(5)表(5)

计量

PDF下载量: 23
文章访问数: 852
HTML全文浏览量: 439
引证文献数: 0

通讯作者: 陈斌, bchen63@163.com

1.
沈阳化工大学材料科学与工程学院沈阳 110142

基于流形学习能量数据预处理的模板攻击优化方法

通讯作者: 王永娟,pinkywyj@163.com

English

An Improved Template Analysis Method Based on Power Traces Preprocessing with Manifold Learning

Corresponding author: Yongjuan WANG,pinkywyj@163.com

计量

文章相关

通讯作者: 陈斌, bchen63@163.com