Bionetta: Efficient Client-Side Zero-Knowledge Machine Learning Proving (arxiv.org)

The paper implies the work could be used for a "biometric proximity proof"; is that right? How would it work?

Minor: the references for the paper are formatted in a careless way