密码学报（中英文）2025，Vol.12Issue(5)：1144-1161,18.DOI:10.13868/j.cnki.jcr.000815
基于神经网络近似结构的PUF建模复杂度分析方法
Modeling Complexity Analysis Method for PUFs Based on Neural Network Approximation Structure
摘要
Abstract
Physical unclonable functions(PUFs)exploit the inherent randomness introduced during the manufacturing process to provide intrinsic security for integrated circuits.Nevertheless,PUFs are susceptible to machine learning and side-channel modeling attacks,making modeling resistance a primary goal in current PUF design.Traditional methods for evaluating the anti-modeling capabilities of PUFs suffer from issues such as the inability to evaluate resistance to side-channel attacks and dis-crepancies between evaluation results and experimental tests.To elucidate the relationship between PUF structure and its ability to accurately resist modeling attacks,a method is proposed for analyz-ing the modeling complexity of PUFs based on neural network models,starting from the modeling capabilities required by adversaries.The proposed approach is based on the universal approximation theorem and the computational complexity theory for neural networks.It generates a neural network model based on the constituent components and connectivity of the targeted PUF to simulate the response behavior and side-channel models.By analyzing the computational complexity metrics of the neural network,the complexity of modeling training for the PUF is derived.The experiments are also conducted to analyze and validate the modeling complexity of XOR arbiter PUF,MPUF,and iPUF by the proposed method.The experimental results demonstrate that the proposed method accurately reflects the difficulty of machine learning and side-channel modeling for PUFs,which is consistent with empirical evaluation results.关键词
物理不可克隆函数/神经网络/近似结构/建模复杂度Key words
physical unclonable function/neural network/approximation structure/modeling com-plexity分类
计算机与自动化引用本文复制引用
刘威,谷大武..基于神经网络近似结构的PUF建模复杂度分析方法[J].密码学报（中英文）,2025,12(5):1144-1161,18.基金项目
上海市科技行动计划(22511101300) （22511101300）
河南省重点研发计划(221111210300)Shanghai Science and Technology Innovation Action Program(22511101300) （221111210300）
Henan Key Research and Development Program(221111210300) （221111210300）
