A guide for assessing optically imaged physically unclonable functions for authentication

Physically unclonable functions (PUFs) are a physical security primitive with important applications in authentication, such as in anti-counterfeiting technologies. They can be used to generate unique identities, linked to their structure, by measuring features associated with them. Optically read PUFs (O-PUFs) are a subset that utilizes optical imaging techniques to create these database of identities. They offer a promising solution to the perpetual challenge of counterfeiting by providing a robust authentication solution. The metrics chosen for evaluation are varied across the field; there is a lack of consensus in the figures of merit used for evaluation, as well as the protocols and standards used for assessing this specific subset of PUFs. This work reviews the progress in the development of evaluation techniques to date, and it highlights important differences in the statistical assessment of O-PUFs. A summary of the most popular metrics used in the literature in the past decade is presented, and the core metrics are isolated and mathematically defined. These are then distilled into recommendations of best practice for assessing and comparing different technologies. An open-source package, providing a full testing suite, is presented to standardize testing in this field. Finally, novel methods for evaluating the performance of O-PUFs over time are also proposed. A unified approach to assessment is essential for advancing anti-counterfeiting technologies, especially as these systems are now being used in commercial applications.