Synapse‐Integrated Physical Unclonable Functions for Trusted Inference in Neuromorphic Computing
Advanced Functional Materials · 05.11.2026
DOI: 10.1002/adfm.79361 ↗Kurz gefasst
Fachartikel in der Zeitschrift Advanced Functional Materials, online erschienen am 05.11.2026. Verfasst von Min Wook Kang und 6 weiteren Autor:innen. Die Arbeit nennt 44 Literaturangaben.
Abstract (Originalsprache)
Secure neuromorphic computing requires compact hardware primitives that can authenticate devices and protect synaptic states without relying on stored keys or large physical arrays. Here, we propose a noise‐spectroscopy physical unclonable function (NS PUF) based on a compact floating‐body (FB) transistor with a charge‐trap layer (CTL). In this platform, the low‐frequency noise power spectral density serves as a frequency‐domain fingerprint of stochastic FB dynamics, while the CTL provides programmable device states for reconfigurable synaptic security. Device geometry diversifies the spectral response landscape, enabling multiple device‐specific responses to be generated from the same compact physical array. The spectral responses are converted into bitmap challenge‐response pairs and optimized to suppress redundant readouts and improve response diversity. The NS PUF‐derived keys can be used to lock or reconfigure synaptic weight‐mapping rules based on bipartite‐sort coding, linking intrinsic device entropy with trusted inference in secure neuromorphic operation. By assigning new security roles to two technologies already compatible with modern microelectronics, FB dynamics and CTL programmability, this work establishes a scalable hardware foundation for secure neuromorphic computing.
Zitieren
APA
Kang, M. W., Park, S., Cho, Y., Baek, S. H., Radermacher, D., Han, J., & Shin, W. (2026). Synapse‐Integrated Physical Unclonable Functions for Trusted Inference in Neuromorphic Computing. Advanced Functional Materials. https://doi.org/10.1002/adfm.79361
Vancouver
Kang MW, Park S, Cho Y, Baek SH, Radermacher D, Han J, et al. Synapse‐Integrated Physical Unclonable Functions for Trusted Inference in Neuromorphic Computing. Advanced Functional Materials. 2026. doi:10.1002/adfm.79361
BibTeX
@article{Kang2026,
author = {Kang, Min Wook and Park, Seonjae and Cho, Youngchan and Baek, Seo Hui and Radermacher, David and Han, Joon‐Kyu and Shin, Wonjun},
title = {Synapse‐Integrated Physical Unclonable Functions for Trusted Inference in Neuromorphic Computing},
journal = {Advanced Functional Materials},
year = {2026},
doi = {10.1002/adfm.79361}
}
EUVIMED