Unclonable MXene Topographies as Robust Anti‐Counterfeiting Tags via Fast Laser Scanning and Siamese Neural Networks

作者：Lin Jing, Hongxiang Si, Tianle Chen, Li‐Yin Hsiao, Haochen Yang, Joshua M. Little, Kerui Li, Shuo Li, Qian Xie, Po‐Yen Chen · 发表于：Advanced Materials Technologies · 年份：2023 · DOI：10.1002/admt.202300568 · 被引用次数：10 · 研究领域：Physical Unclonable Functions (PUFs) and Hardware Security、Advanced Memory and Neural Computing、Neuroscience and Neural Engineering

Abstract An ideal anti‐counterfeiting technology is desired to be unclonable, nondestructive, mass‐producible, and accompanied with fast and robust authentication under various external influences. Although multiple anti‐counterfeiting technologies have been reported, few meet all of the above‐mentioned features. Herein, a mechanically driven patterning process is reported to produce higher dimensional Ti 3 C 2 T x MXene topographies in a scalable yet unclonable manner, which can be used as anti‐counterfeiting tags. By using a high‐speed confocal laser microscopy, the complex topographies can be extracted within one minute and then reconstructed into 3D physical unclonable function (PUF) keys. Meanwhile, a Siamese neural network model and a feature‐tracking software are built to achieve a pick‐and‐check strategy, enabling highly accurate, robust, disturbance‐insensitive tag authentication in practical exploitations. The 3D PUF key‐based anti‐counterfeiting technology features with several advances, including ultrahigh encoding capacities (≈10 144 000‐ 10 7 800 000 ), fast processing times (<1 min), and high authentication accuracy under various external disturbances, including tag rotations (≈0°‒360°), tag dislocation(s) in x ( y ) directions (≈0%‒100%), tag shifts in z ‐direction (≈0%‒28%), tag tilts (≈0°‒5°), differences in contrasts (20%‒60%) and laser power (6.0‒9.0 µW). The anti‐counterfeiting technology promises information security, encoding capacity, and authentica...