Programmable Physical Unclonable Functions Using Randomly Anisotropic Two-Dimensional Flakes

Physical unclonable functions (PUFs) have been developed as promising strategies for secure authentication. Conventional strategies of PUFs have a limitation in the aspect of security for their static single channel. The introduction of polarization will endow a static PUF with many dynamic transformations based on polarized properties. Herein, high-security PUFs based on the polarized luminescence of chaotic luminescent patterns are fabricated by anisotropic rare earth (RE) material Er3O4Cl flakes. These derivatives under different polarizations show strong randomness (with similarity of the same PUF as high as 97%, while that for different PUFs is as low as 44%), which further widens the security and capacity of PUFs. Polarized luminescence control of Er3O4Cl patterns gives freedom to the PUFs and ensures a high encoding capacity of 2(380000). Furthermore, we build a convolutional neural network (CNN) to realize fast intelligent authentication by extracting image features for convolution operation with a very high accuracy of 99.8% and fast classification speed in only 5 epochs.