Two-Stage Architectures for Resilient Lightweight PUFs

Abstract

A Physical Unclonable Function provides a secure method for lightweight device authentication, in contrast to traditional encryption methods that are expensive in terms of resources. Unfortunately, they have proved to be vulnerable to modelling attacks. This work shows that two-stage structures improve the resilience against machine learning attacks. We have evaluated a number of two-stage structures using combinations of Arbiter PUFs (APUF), Current Mirror PUFs (CM-PUF) and Differential Comparator PUFs (DC-PUF). Of these, a DC-DC-PUF with an XOR at the output has a predictability of about 50% Additionally, it has a high reliability of 95.2% and only requires 26.1% of the area of the best existing approach.