Polarization-direction multiplexed metasurfaces for optical information hiding and decryption

The continuous development of information and communication technologies has made information hiding and decryption a rapidly growing of interest. Optical information offers multi-dimensional characteristics and high response speeds as a means of carrying data. However, conventional methods for hiding and decrypting optical information require redundant optical components or high design capacity, leading to inefficiency. The emergence of the diffractive deep neural network (D²NN) introduces a novel approach. Metasurfaces, two-dimensional planar structures with unique electromagnetic responses, can flexibly manipulate the optical field at subwavelength scales, providing an efficient platform for optical information hiding and decryption. Here, we propose polarization-direction multiplexed metasurfaces for optical information hiding and decryption, trained via D²NN to correlate input with optical information, and integrating phases by utilizing the polarization- and direction-sensitive properties of the metasurfaces. This method introduces a new design concept and approach for hiding and decrypting optical information.