【喜报】本课题组成员在IEEE Electron Device Letters发表文章一篇
29 November 2021
               

2021年11月22日,本课题组涂泽中、薛永康两位同学合作在微电子领域权威期刊IEEE Electron Device Letters(影响因子4.187)发表题为“A Probability-based Strong Physical Unclonable Function with Strong Machine Learning Immunity”文章,纪志罡教授为通讯作者。

近年来,在机器学习高速发展的大背景下,基于软件方法生成的密钥容易被黑客攻击,可靠程度逐渐降低,已无法满足当前的安防需求。为增强系统的可靠性,利用固体器件在掺杂、老化等过程中出现随机性差异所设计的物理不可克隆函数(Physical unclonable function,PUF)逐渐为越来越多的安防系统所需求。

这篇文章基于晶体管栅氧化层缺陷对沟道载流子捕获与释放的随机行为,提出了一种新的强物理不可克隆函数——基于概率的PUF(Prob-PUF)。首次将缺陷释放载流子的概率信息应用于PUF设计中,显著提升了PUF系统对机器学习的免疫能力。同时提出的以概率模型存储激励响应对(CRP)的方式,解决了大量CRP需求和有限存储空间之间的困境,为未来安全存储提供了潜在的解决方案。

Abstract:

A novel strong physical unclonable function(PUF), called Probability-based PUF (Prob-PUF), is proposed using the stochastic process of trap emission in nano-scaled transistors. For the first time, the information of trap emission probability is used in the PUF design. This new approach offers ideal immunity to machine learning (ML) attacks. Since Prob-PUF only stores a mathematical model, it naturally avoids the dilemma between the requirement of a large number of challenge-response pairs (CRPs) and the limited storage space, making it a potential solution for future secure storage.