Physical Unclonable Functions From Microwave Metasurface Fingerprints

ABSTRACT Electromagnetic (EM) metasurfaces enable precise control over wave amplitude, phase, and polarization, and are increasingly used in wireless systems for communications and sensing. As these metasurface‐enabled systems move toward practical deployment, reliable wireless authentication becomes essential for preventing device spoofing and unauthorized access. This work proposes a metasurface‐based physical unclonable function (MetaPUF) that endows a functional metasurface with an additional authentication capability beyond its normal wavefront‐control operation. The key principle is that nominally identical metasurfaces fabricated through the same manufacturing process retain the designed macroscopic wave‐control function, but inevitably contain hardware‐level imperfections during lithography, etching, and material processing. These imperfections cause subtle, device‐specific deviations from the nominal EM response. A neural network extracts these deviations as EM fingerprints, represented by discriminative feature embeddings, for device authentication. The proposed framework supports both closed‐set identification of enrolled metasurfaces and open‐set rejection of previously unseen devices. Experiments on ten fabricated 1‐bit metasurfaces achieve 94.29% closed‐set identification accuracy at 3 dB SNR. For open‐set authentication, Normalized IQ increases the ROC‐AUC from 0.8412 to 0.9011 relative to the baseline. These results demonstrate that ordinary functional metasurfaces can serve as self‐authenticating EM hardware primitives by exploiting fabrication imperfections that are difficult to avoid in practical manufacturing.