基于对抗学习的强PUF安全结构研究

doi:10.11959/j.issn.2096-109x.2021019

网络与信息安全学报 ›› 2021, Vol. 7 ›› Issue (3): 115-122.doi: 10.11959/j.issn.2096-109x.2021019

基于对抗学习的强PUF安全结构研究

李艳, 刘威, 孙远路

信息工程大学，河南郑州 450001

修回日期:2020-07-02 出版日期:2021-06-15 发布日期:2021-06-01
作者简介:李艳（1988- ）女，河南郑州人，信息工程大学讲师，主要研究方向为网络空间安全
刘威（1982- ）男，湖北随州人，信息工程大学讲师，主要研究方向为硬件安全
孙远路（1974- ）男，河南商丘人，信息工程大学副教授，主要研究方向为管理科学与工程
基金资助:
国家自然科学基金(61871405);国家自然科学基金(61802431)

Research on security architecture of strong PUF by adversarial learning

Yan LI, Wei LIU, Yuanlu SUN

Information Engineering University, Zhengzhou 450001, China

Revised:2020-07-02 Online:2021-06-15 Published:2021-06-01
Supported by:
The National Natural Science Foundation of China(61871405);The National Natural Science Foundation of China(61802431)

摘要/Abstract

摘要：

针对强物理不可复制函数（PUF，physical unclonable function）面临的机器学习建模威胁，基于对抗学习理论建立了强PUF的对抗机器学习模型，在模型框架下，通过对梯度下降算法训练过程的分析，明确了延迟向量权重与模型预测准确率之间的潜在联系，设计了一种基于延迟向量权重的对抗样本生成策略。该策略与传统的组合策略相比，将逻辑回归等算法的预测准确率降低了5.4%~9.5%，低至51.4%。结合资源占用量要求，设计了新策略对应的电路结构，并利用对称设计和复杂策略等方法对其进行安全加固，形成了ALPUF（adversarial learning PUF）安全结构。ALPUF不仅将机器学习建模的预测准确率降低至随机预测水平，而且能够抵御混合攻击和暴力破解。与其他 PUF结构的对比表明，ALPUF在资源占用量和安全性上均具有明显优势。

关键词: 物理不可复制函数, 对抗样本, 延迟向量, 对抗学习PUF

Abstract:

To overcome the vulnerability of strong physical unclonable function, the adversarial learning model of strong PUF was presented based on the adversarial learning theory, then the training process of gradient descent algorithm was analyzed under the framework of the model, the potential relationship between the delay vector weight and the prediction accuracy was clarified, and an adversarial sample generation strategy was designed based on the delay vector weight.Compared with traditional strategies, the prediction accuracy of logistic regression under new strategy was reduced by 5.4% ~ 9.5%, down to 51.4%.The physical structure with low overhead was designed corresponding to the new strategy, which then strengthened by symmetrical design and complex strategy to form a new PUF architecture called ALPUF.ALPUF not only decrease the prediction accuracy of machine learning to the level of random prediction, but also resist hybrid attack and brute force attack.Compared with other PUF security structures, ALPUF has advantages in overhead and security.

Key words: strong physical unclonable function, adversarial sample, delay vector, adversarial learning PUF

中图分类号:

TP393

李艳, 刘威, 孙远路. 基于对抗学习的强PUF安全结构研究[J]. 网络与信息安全学报, 2021, 7(3): 115-122.

Yan LI, Wei LIU, Yuanlu SUN. Research on security architecture of strong PUF by adversarial learning[J]. Chinese Journal of Network and Information Security, 2021, 7(3): 115-122.

图/表 9

图1

表1

图2

表2

表3

图3

图4

图5

表4

参考文献 16

[1]	SAHOO D P , NGUYEN P H , MUKHOPADHYAY D ,et al. A case of lightweight PUF constructions:cryptanalysis and machine learning attacks[J]. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2015,34(8): 1334-1343.
[2]	LIU Y , XIE Y , BAO C ,et al. A combined optimization-theoretic and side-channel approach for attacking strong physical unclonable functions[J]. IEEE Transactions on Very Large Scale Integration Systems, 2017,26(1): 73-81.
[3]	RüHRMAIR U , S?LTER J , SEHNKE F ,et al. PUF modeling attacks on simulated and silicon data[J]. IEEE Transactions on Information Forensics ＆ Security, 2013,8(11): 1876-1891.
[4]	YE J , GUO Q , HU Y ,et al. Modeling attacks on strong physical unclonable functions strengthened by random number and weak PUF[C]// 2018 IEEE 36th VLSI Test Symposium. 2018: 1-6.
[5]	DELVAUX J . Machine-learning attacks on PolyPUFs,OB-PUFs,RPUFs,LHS-PUFs,and PUF–FSMs[J]. IEEE Transactions on Information Forensics and Security, 2019,14(8): 2043-2058.
[6]	宋蕾, 马春光, 段广晗 . 机器学习安全及隐私保护研究进展[J]. 网络与信息安全学报, 2018,4(8): 1-11.
	SONG L , MA C G , DUAN G H . Machine learning security and privacy:a survey[J]. Chinese Journal of Network and Information Security, 2018,4(8): 1-11.
[7]	BIGGIO B , NELSON B , LASKOV P . Poisoning attacks against support vector machines[C]// ICML'12 2012. 14671-1474.
[8]	JAGIELSKI M , OPREA A , BIGGIO B ,et al. Manipulating machine learning:poisoning attacks and countermeasures for regression learning[C]// 2018 IEEE Symposium on Security and Privacy. 2018. 19-35.
[9]	BIGGIO B , ROLI F . Wild patterns:ten years after the rise of adversarial machine learning[J]. Pattern Recognition, 2018,84: 317-331.
[10]	BECKER G T , . The gap between promise and reality:on the insecurity of XOR arbiter PUFs[C]// Cryptographic Hardware and Embedded Systems. 2015: 535-555.
[11]	WANG S J , CHEN Y S , KATHERINE L ,et al. Adversarial attack against modeling attack on PUFs[C]// DAC2019. 2019: 1-6.
[12]	PRADO H , PATIL C . Defeating strong PUF modeling attack via adverse selection of challenge-response pairs[C]// AsianHOST2018. 2018: 25-30.
[13]	XU X , MAHMOUD A , MAJZOOBI M ,et al. Efficient power and timing side channels for physical unclonable functions[C]// International Workshop on Cryptographic Hardware and Embedded Systems. 2014: 476-492.
[14]	GAO Y , LI G , MA H ,et al. Obfuscated challenge-response:a secure lightweight authentication mechanism for PUF-based pervasive devices[C]// Workshop in Conjunction with IEEE International Conference on Pervasive Computing and Communications. 2016: 1-6.
[15]	JING Y , YU H , LI X . RPUF:physical unclonable function with randomized challenge to resist modeling attack[C]// Hardware-oriented Security ＆ Trust. 2016: 1-6.
[16]	徐金甫, 吴缙, 李军伟 . 基于敏感度混淆机制的控制型物理不可复制函数研究[J]. 电子与信息学报, 2019,41(7): 1601-1609.
	XU J F , WU J , LI J W . Research on the control physical non-clonable function based on sensitivity cofounding mechanism[J]. Journal of Electronics and Information Technology, 2019,41(7): 1601-1609.

W取值	预测准确率
W取值	LR	SVM
-0.021 266 21 （Min）	51.4%	51.7%
2.323 293 36（Max）	62.3%	62.5%
1.042 470 09（Avg）	56.65%	56.9%
Nul	99.8%	99%

策略	预测准确率
策略	LR	SVM
本文	51.4%	51.7%
随机翻转^[11]	50.2%	49.7%
组合翻转1^[11]	56.8%	57.4%
组合翻转2^[11]	60.9%	61.3%

投毒比例	预测准确率
投毒比例	基于延迟向量权重	组合翻转
0%	99.8%	99.8%
12.5%	86.02%	90.6%
25%	73%	78.2%
50%	51.4%	56.8%

类型	预测准确率（基于LR）	对抗功率混合攻击	资源占用量
AL PUF	51.4%	√	3 XOR
4-XOR PUF	98.2%	×	3 XOR 3 APUF
OB PUF^[14]	72%	√	1 TRNG3 APUF
R PUF^[15]	80%	√	1 TRNG 1 LFSR
控制型PUF^[16]	69.8%	√	1 XOR 1 TRNG 1混淆模块

基于对抗学习的强PUF安全结构研究

Research on security architecture of strong PUF by adversarial learning

在线阅读

pdf下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 16

相关文章 7

Metrics

推荐阅读 0

[1]	林点, 潘理, 易平. 面向图像识别的卷积神经网络鲁棒性研究进展[J]. 网络与信息安全学报, 2022, 8(3): 111-122.
[2]	邱宝琳, 易平. 基于多维特征图知识蒸馏的对抗样本防御方法[J]. 网络与信息安全学报, 2022, 8(2): 88-99.
[3]	秦中元, 贺兆祥, 李涛, 陈立全. 基于图像重构的MNIST对抗样本防御算法[J]. 网络与信息安全学报, 2022, 8(1): 86-94.
[4]	王滨, 陈靓, 钱亚冠, 郭艳凯, 邵琦琦, 王佳敏. 面向对抗样本攻击的移动目标防御[J]. 网络与信息安全学报, 2021, 7(1): 113-120.
[5]	刘西蒙,谢乐辉,王耀鹏,李旭如. 深度学习中的对抗攻击与防御[J]. 网络与信息安全学报, 2020, 6(5): 36-53.
[6]	段广晗,马春光,宋蕾,武朋. 深度学习中对抗样本的构造及防御研究[J]. 网络与信息安全学报, 2020, 6(2): 1-11.
[7]	严飞,张铭伦,张立强. 基于边界值不变量的对抗样本检测方法[J]. 网络与信息安全学报, 2020, 6(1): 38-45.