Machine Learning Resistant Amorphous Silicon Physically Unclonable Functions (PUFs)

• DOI: 10.48550/arxiv.2402.02846
• Publication Date: 2024-02-05
• Authors: Velat Kilic; Neil MacFarlane; Jasper Stround; Samuel Métais; Milad Alemohammad; A. Brinton Cooper; Amy C. Foster; Mark A. Foster
• Keywords: Physical unclonable function

We investigate usage of nonlinear wave chaotic amorphous silicon (a-Si) cavities as physically unclonable functions (PUF). Machine learning attacks on integrated electronic PUFs have been demonstrated to be very effective at modeling PUF behavior. Such a-Si photonic are investigated through application algorithms including linear regression, k-nearest neighbor, decision tree ensembles (random forests and gradient boosted trees), deep neural networks (DNNs). found that DNNs performed the best among all studied but still failed completely break security which we quantify a private information metric. Furthermore, machine resistance were directly related strength their response.