Physical unclonable functions (PUFs) offer a different strategy: instead of storing a secret, they use tiny and unavoidable manufacturing variations in a physical object to generate a unique response.
Optical PUFs use light to read these physical variations. When light interacts with a complex material or optical structure, the resulting pattern can be extremely difficult to reproduce exactly.
Now a research team, headed by Prof. Seong Chan Jun from School of Mechanical Engineering at Yonsei University in South Korea, addressed this challenge using graphene diffractive zone plates. A diffractive zone plate is an ultrathin optical element that focuses light through interference rather than through the curved surfaces used in conventional lenses.
Different Colors
Because diffraction depends strongly on wavelength, the same graphene zone plate responds differently when illuminated with different combinations of red, green, and blue light.
In the proposed system, the color composition of the incident light acts as an optical challenge. The graphene diffractive zone plate then produces a wavelength-dependent focal and interference pattern, which serves as the physical response. These patterns contain information from the designed Fresnel-zone geometry and small fabrication-induced variations in the multilayer graphene structure.
To make the optical response usable for authentication, researchers used a vision transformer model to analyze the captured diffraction images.
The model converts each high-dimensional RGB diffraction pattern into a compact numerical representation. Stable components of this representation are then selected and converted into binary response units that can end up compared using standard authentication metrics.
This approach turns a complex optical image into a repeatable digital response while preserving the physical uniqueness and challenge dependence of the optical system.
RGB spectral challenges are converted by graphene diffractive zone plates into wavelength-dependent focal and interference patterns, which are analyzed by a transformer-based readout model to generate compact binary response units for authentication.
Source: Prof. Seong Chan Jun from Yonsei University
Rather than relying on a single fixed image, the platform can generate different responses under different spectral challenges. In this way, the work demonstrates a spectrally programmable diffractive optical PUF that bridges physical optics, graphene nanomaterials, and artificial intelligence-based readout.
This study contributes to the broader effort to build security functions directly into physical hardware. As artificial intelligence, edge computing, and connected devices become more widespread, authentication cannot rely only on software credentials or stored digital secrets. Hardware-level identifiers are difficult to copy or predict can provide an additional layer of trust for future electronic systems.
Combines Two Properties
The proposed platform is important because it combines two properties rarely achieved together in optical PUFs. First, the graphene diffractive zone plate provides a physically complex optical response shaped by fabrication-induced local heterogeneity. Second, the response can end up reconfigured by changing the RGB spectral challenge without mechanically moving the device or scanning the beam position. This enables a fixed optical geometry to support multiple challenge-response states.
Another significance of the work lies in the conversion of analog optical patterns into compact binary response units.
Optical systems naturally generate rich images, but practical authentication protocols require stable digital outputs that can end up compared, stored, and verified. By using a vision transformer to extract global diffraction features and selecting reliable embedding dimensions for binarization, the team demonstrates a pathway from optical complexity to authentication-oriented digital responses.
The approach may be useful for future optical security tags, anti-counterfeiting technologies, chip or package authentication, and hardware roots of trust. In such applications, the optical structure would act as a physical layer that is difficult to clone, while the spectral challenge sequence would provide a programmable way to interrogate the device.
The work should be understood as a proof-of-concept demonstration rather than a finalized manufacturing platform.
The current graphene transfer and focused-ion-beam patterning process ended up used to accurately define the diffractive structures and validate the principle.
Future Versions
Future implementations may explore scalable fabrication routes, integrated optical readers, and chip- or package-level authentication modules.
More broadly, the study shows how nanophotonic hardware and learning-based analysis can work together in security systems. Instead of treating artificial intelligence only as a classifier, the model can stabilize and interpret physically meaningful optical responses. This combination may inspire new forms of programmable optical authentication, secure photonic hardware, and adaptive physical security systems.
Click here to review the paper on the subject.