The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
Este artigo propõe uma rede neural profunda chamada BayesianaPUFNet que pode alcançar alta precisão de previsão mesmo com poucos pares desafio-resposta (CRPs) disponíveis para treinamento. Geralmente, os ataques de modelagem são uma vulnerabilidade que pode comprometer a autenticidade de funções fisicamente não clonáveis (PUFs); assim, vários métodos de aprendizado de máquina, incluindo redes neurais profundas, foram propostos para avaliar a vulnerabilidade dos PUFs. No entanto, os ataques de modelagem convencional não consideraram o custo da coleta de CRP e analisaram os ataques com base na suposição de que CRPs suficientes estavam disponíveis para treinamento; portanto, estudos anteriores podem ter subestimado a vulnerabilidade dos PUFs. Aqui, mostramos que a aplicação de redes neurais profundas Bayesianas que incorporam estatísticas Bayesianas pode fornecer previsão precisa de respostas mesmo em situações onde CRPs suficientes não estão disponíveis para aprendizagem. Experimentos numéricos mostram que o modelo proposto utiliza apenas metade da PCR para obter a mesma previsão de resposta dos métodos convencionais. Nosso código está disponível abertamente em https://github.com/bayesian-puf-net/bayesian-puf-net.git.
Hiromitsu AWANO
Kyoto University
Makoto IKEDA
The University of Tokyo
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.
Copiar
Hiromitsu AWANO, Makoto IKEDA, "BayesianPUFNet: Training Sample Efficient Modeling Attack for Physically Unclonable Functions" in IEICE TRANSACTIONS on Fundamentals,
vol. E106-A, no. 5, pp. 840-850, May 2023, doi: 10.1587/transfun.2022EAP1061.
Abstract: This paper proposes a deep neural network named BayesianPUFNet that can achieve high prediction accuracy even with few challenge-response pairs (CRPs) available for training. Generally, modeling attacks are a vulnerability that could compromise the authenticity of physically unclonable functions (PUFs); thus, various machine learning methods including deep neural networks have been proposed to assess the vulnerability of PUFs. However, conventional modeling attacks have not considered the cost of CRP collection and analyzed attacks based on the assumption that sufficient CRPs were available for training; therefore, previous studies may have underestimated the vulnerability of PUFs. Herein, we show that the application of Bayesian deep neural networks that incorporate Bayesian statistics can provide accurate response prediction even in situations where sufficient CRPs are not available for learning. Numerical experiments show that the proposed model uses only half the CRP to achieve the same response prediction as that of the conventional methods. Our code is openly available on https://github.com/bayesian-puf-net/bayesian-puf-net.git.
URL: https://global.ieice.org/en_transactions/fundamentals/10.1587/transfun.2022EAP1061/_p
Copiar
@ARTICLE{e106-a_5_840,
author={Hiromitsu AWANO, Makoto IKEDA, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={BayesianPUFNet: Training Sample Efficient Modeling Attack for Physically Unclonable Functions},
year={2023},
volume={E106-A},
number={5},
pages={840-850},
abstract={This paper proposes a deep neural network named BayesianPUFNet that can achieve high prediction accuracy even with few challenge-response pairs (CRPs) available for training. Generally, modeling attacks are a vulnerability that could compromise the authenticity of physically unclonable functions (PUFs); thus, various machine learning methods including deep neural networks have been proposed to assess the vulnerability of PUFs. However, conventional modeling attacks have not considered the cost of CRP collection and analyzed attacks based on the assumption that sufficient CRPs were available for training; therefore, previous studies may have underestimated the vulnerability of PUFs. Herein, we show that the application of Bayesian deep neural networks that incorporate Bayesian statistics can provide accurate response prediction even in situations where sufficient CRPs are not available for learning. Numerical experiments show that the proposed model uses only half the CRP to achieve the same response prediction as that of the conventional methods. Our code is openly available on https://github.com/bayesian-puf-net/bayesian-puf-net.git.},
keywords={},
doi={10.1587/transfun.2022EAP1061},
ISSN={1745-1337},
month={May},}
Copiar
TY - JOUR
TI - BayesianPUFNet: Training Sample Efficient Modeling Attack for Physically Unclonable Functions
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 840
EP - 850
AU - Hiromitsu AWANO
AU - Makoto IKEDA
PY - 2023
DO - 10.1587/transfun.2022EAP1061
JO - IEICE TRANSACTIONS on Fundamentals
SN - 1745-1337
VL - E106-A
IS - 5
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - May 2023
AB - This paper proposes a deep neural network named BayesianPUFNet that can achieve high prediction accuracy even with few challenge-response pairs (CRPs) available for training. Generally, modeling attacks are a vulnerability that could compromise the authenticity of physically unclonable functions (PUFs); thus, various machine learning methods including deep neural networks have been proposed to assess the vulnerability of PUFs. However, conventional modeling attacks have not considered the cost of CRP collection and analyzed attacks based on the assumption that sufficient CRPs were available for training; therefore, previous studies may have underestimated the vulnerability of PUFs. Herein, we show that the application of Bayesian deep neural networks that incorporate Bayesian statistics can provide accurate response prediction even in situations where sufficient CRPs are not available for learning. Numerical experiments show that the proposed model uses only half the CRP to achieve the same response prediction as that of the conventional methods. Our code is openly available on https://github.com/bayesian-puf-net/bayesian-puf-net.git.
ER -