On the Importance of Physical Model Parameters for PUF Performance: A Case Study on BFO Memristors

A common approach when designing a memristorbased Physical Unclonable Function (PUF) is to quantize directly observable electrical quantities. While this is a straightforward approach, it suffers from a problem: It remains unclear which effects underly the extracted entropy. As a consequence, it is hard to counteract detrimental effects, e.g. bias and correlation, due to their unknown causes, and to optimize measurement and quantization techniques. This work provides the foundation to base a PUF on physical model parameters extracted from $I-V$ curve measurements, using the reading branch of a BiFeO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</inf>(BFO) memristor as an example. The work outlines possible avenues for deriving reliable and high-entropy PUF responses from such parameters in general and concretely for BFO memristors using experimental data from a laboratory sample with 129 memristive cells.