Lightweight and Modeling Attack Resistant PUFs Authentication Based on Portion Mapping

Physical Unclonable Functions provide an efficient authentication scheme over traditional security-related fields based on inherent manufacturing variations. While the transmission channel of authentication protocol is exposed where adversaries collect challenge response pairs and machine learning methods are performed on these data, that result in false certification. In this paper, a novel PUFs-based authentication protocol was proposed which has resistance against machine learning attack and keep hardware consumption low-overhead. A response attenuator based on Von Neumann corrector is applied in this protocol to mapping specific portion relation between challenge and response. This scheme provides verifier with inherent noisy CRPs while the same CPRs turn out to be irrelevant to adversaries due to portion mapping of CRPs. We present the implementation on 28nm FPGA as well as ES modeling attack results. This protocol reduces prediction rate of 2-XOR APUF from 75% to 57% with only hundreds of 6-input Look-Up-Tables and registers.