Authors: Velat Kilic, Neil Macfarlane, Jasper Stround, Samuel Metais, Milad Alemohammad, A. Brinton Cooper, Amy C. Foster, Mark A. Foster

Published on: February 05, 2024

Impact Score: 8.22

Arxiv code: Arxiv:2402.02846

Summary

• What is new: The study explores the resistance of amorphous silicon (a-Si) cavities used as physically unclonable functions (PUFs) against machine learning attacks, a novel area of research showing the robustness of a-Si PUFs in secure applications.
• Why this is important: Machine learning attacks on integrated electronic PUFs have been very effective, posing a security risk.
• What the research proposes: Investigating the resistance of integrated a-Si photonic PUFs to machine learning attacks to enhance security.
• Results: Deep neural networks (DNNs) were the most effective in modeling PUF behavior but could not completely compromise a-Si PUF security, indicating their potential in secure applications.

Technical Details

Technological frameworks used: nan

Models used: Linear regression, k-nearest neighbor, decision trees (random forests, gradient boosted trees), deep neural networks (DNNs)

Data used: nan

Potential Impact

Security and encryption markets, particularly companies involved in crafting secure communication and authentication technologies.

Want to implement this idea in a business?

We have generated a startup concept here: SecureSilicon.