Physical Unclonable Functions with Boson Sampling
Abstract
Boson sampling is a problem for which quantum devices could prove to go beyond classical computing using only linear optics and photon preparation and counting. While theoretically important, there is a lack of practical applications for this task. This paper presents a way to turn any boson sampling device for which there are certain fabrication limits into a physically unclonable function which could be used in identification systems and other security protocols. For a minimum fabrication error, there are scenarios for which the photon distribution at the output of two different devices has a minimum distance and they can be told apart. When compared to classical PUFs these linear optical tokens have the advantage that, for the right choice of parameters, there is a computational guarantee that no classical computer can simulate the output. The paper describes the basic limitations of boson sampling which allow these results and gives a possible choice for a challenge-response identification protocol. This gives a direct application of boson sampling which could be simpler than solving the general problem.
Cite
@article{arxiv.1911.08417,
title = {Physical Unclonable Functions with Boson Sampling},
author = {Juan Carlos Garcia-Escartin},
journal= {arXiv preprint arXiv:1911.08417},
year = {2019}
}
Comments
The paper has two flaws that make the scheme useless. As noted by many correspondents, two samples from the same device cannot be directly compared. There are exponentially many sampled values and the comparison is inefficient. This could be solved with some binning methods mapping many samples to one bin, but then the distance to the distributions of imperfect clones is not clear. The second problem is that there are efficient ways to produce samples classically which cannot be told apart from the correct ones with a reasonable number of steps