A Systematic Method to Evaluate and Compare the Performance of Physical Unclonable Functions

Abstract

We propose a systematic method to evaluate and compare the performance of physical unclonable functions (PUFs). The need for such a method is justified by the fact that various types of PUFs have been proposed so far. However, there is no common method that can fairly compare them in terms of their performance. We first propose three generic dimensions of PUF measurement. We then define several parameters to quantify the performance of a PUF along these dimensions. We also analyze existing parameters proposed by other researchers. Based on our analysis, we propose a compact set of parameters that will be used as a tool to evaluate as well as compare the performance of different PUFs. To make the method independent of the underlying PUF technique, we focus on the statistical properties of the binary PUF responses. We demonstrate the proposed method with a detailed comparison analysis between two PUFs: the ring-oscillator-based PUF (RO PUF) and the Arbiter-based PUF (APUF) using measured data from PUF implementations in state-of-the-art FPGAs. Finally, we present an online database where our measurements and analysis results can be consulted. Our dataset comprises measurements in 193 FPGAs.