Side-channel analysis is one of the most investigated hardware Trojan detection approaches. However, nearly all the side-channel analysis approaches require golden chips for reference, which are hard to obtain actually. Besides, majority of existing Trojan detection algorithms focus on the data similarity and ignore the Trojan misclassification during the detection. In this paper, we propose a cost-sensitive golden chip-free hardware Trojan detection framework, which aims to minimize the probability of Trojan misclassification during the detection. The post-layout simulation data of voltage variations at different process corners is utilized as a golden reference. Further, a classification algorithm based on the combination of principal component analysis and Naïve bayes is exploited to identify the existence of hardware Trojan with a minimum misclassification risk. Experimental results on ASIC demonstrate that the proposed approach improves the detection accuracy ratio compared with the three detection algorithms and distinguishes the Trojan with only 0.27% area occupies even under ±15% process variations.
Yanjiang LIU
the Information Engineering University
Xianzhao XIA
the Tianjin University and China Automotive Technology and Research Center
Jingxin ZHONG
the Information Engineering University
Pengfei GUO
the Information Engineering University
Chunsheng ZHU
the Information Engineering University
Zibin DAI
the Information Engineering University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copy
Yanjiang LIU, Xianzhao XIA, Jingxin ZHONG, Pengfei GUO, Chunsheng ZHU, Zibin DAI, "A Cost-Sensitive Golden Chip-Free Hardware Trojan Detection Using Principal Component Analysis and Naïve Bayes Classification Algorithm" in IEICE TRANSACTIONS on Fundamentals,
vol. E105-A, no. 6, pp. 965-974, June 2022, doi: 10.1587/transfun.2021EAP1060.
Abstract: Side-channel analysis is one of the most investigated hardware Trojan detection approaches. However, nearly all the side-channel analysis approaches require golden chips for reference, which are hard to obtain actually. Besides, majority of existing Trojan detection algorithms focus on the data similarity and ignore the Trojan misclassification during the detection. In this paper, we propose a cost-sensitive golden chip-free hardware Trojan detection framework, which aims to minimize the probability of Trojan misclassification during the detection. The post-layout simulation data of voltage variations at different process corners is utilized as a golden reference. Further, a classification algorithm based on the combination of principal component analysis and Naïve bayes is exploited to identify the existence of hardware Trojan with a minimum misclassification risk. Experimental results on ASIC demonstrate that the proposed approach improves the detection accuracy ratio compared with the three detection algorithms and distinguishes the Trojan with only 0.27% area occupies even under ±15% process variations.
URL: https://global.ieice.org/en_transactions/fundamentals/10.1587/transfun.2021EAP1060/_p
Copy
@ARTICLE{e105-a_6_965,
author={Yanjiang LIU, Xianzhao XIA, Jingxin ZHONG, Pengfei GUO, Chunsheng ZHU, Zibin DAI, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={A Cost-Sensitive Golden Chip-Free Hardware Trojan Detection Using Principal Component Analysis and Naïve Bayes Classification Algorithm},
year={2022},
volume={E105-A},
number={6},
pages={965-974},
abstract={Side-channel analysis is one of the most investigated hardware Trojan detection approaches. However, nearly all the side-channel analysis approaches require golden chips for reference, which are hard to obtain actually. Besides, majority of existing Trojan detection algorithms focus on the data similarity and ignore the Trojan misclassification during the detection. In this paper, we propose a cost-sensitive golden chip-free hardware Trojan detection framework, which aims to minimize the probability of Trojan misclassification during the detection. The post-layout simulation data of voltage variations at different process corners is utilized as a golden reference. Further, a classification algorithm based on the combination of principal component analysis and Naïve bayes is exploited to identify the existence of hardware Trojan with a minimum misclassification risk. Experimental results on ASIC demonstrate that the proposed approach improves the detection accuracy ratio compared with the three detection algorithms and distinguishes the Trojan with only 0.27% area occupies even under ±15% process variations.},
keywords={},
doi={10.1587/transfun.2021EAP1060},
ISSN={1745-1337},
month={June},}
Copy
TY - JOUR
TI - A Cost-Sensitive Golden Chip-Free Hardware Trojan Detection Using Principal Component Analysis and Naïve Bayes Classification Algorithm
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 965
EP - 974
AU - Yanjiang LIU
AU - Xianzhao XIA
AU - Jingxin ZHONG
AU - Pengfei GUO
AU - Chunsheng ZHU
AU - Zibin DAI
PY - 2022
DO - 10.1587/transfun.2021EAP1060
JO - IEICE TRANSACTIONS on Fundamentals
SN - 1745-1337
VL - E105-A
IS - 6
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - June 2022
AB - Side-channel analysis is one of the most investigated hardware Trojan detection approaches. However, nearly all the side-channel analysis approaches require golden chips for reference, which are hard to obtain actually. Besides, majority of existing Trojan detection algorithms focus on the data similarity and ignore the Trojan misclassification during the detection. In this paper, we propose a cost-sensitive golden chip-free hardware Trojan detection framework, which aims to minimize the probability of Trojan misclassification during the detection. The post-layout simulation data of voltage variations at different process corners is utilized as a golden reference. Further, a classification algorithm based on the combination of principal component analysis and Naïve bayes is exploited to identify the existence of hardware Trojan with a minimum misclassification risk. Experimental results on ASIC demonstrate that the proposed approach improves the detection accuracy ratio compared with the three detection algorithms and distinguishes the Trojan with only 0.27% area occupies even under ±15% process variations.
ER -