A Modeling Attack Resistant R-XOR APUF Based on FPGA
Fukui Dan, Yehan Xu, Zheng Li, Jing Wen, Ben Liu, Shuai Chen, Bing Li · 2018
A Physically Unclonable Functions (PUFs) extracts the manufacturing variations of integrated circuits for key generation and authentication. It can be used to address the security issue in traditional non-volatile memory (NVM)-based key generation and authentication system. However, the powerful modeling attack based on machine learning has become a new threat of Strong PUFs-based authentication scheme. In this paper, we proposed a novel reconfigurable XOR Arbiter Physical Unclonable Functions (R-XOR APUFs) to resist this modeling attack. In this paper, the R-XOR APUF consist of multiplexers and inverters. The structure of generating two responses is configured according to challenges. The response of R-XOR APUFs is generated by XORing the two response. Therefore, R-XOR APUFs does not have a uniform model and effectively resists machine learning-based modeling attacks. The experiment results reveal that the uniqueness of R-XOR APUFs is 42.15% (the idea value is 50%) and the prediction rate of R-XOR APUFs reduces from 95% to 55% (the idea value is 50%) compared to the traditional APUFs.