Complex‐Field Diffractive Processor‐Enabled Optical Encryption

ABSTRACT Optical diffractive deep neural networks (D 2 NNs), utilizing light as the medium for all‐optical computing, have paved a new avenue for information security due to their inherent advantages of high speed, low power consumption, and high parallelism. However, most encryption approaches rely solely on the intensity of diffractive processing, yet the security potential of the phase dimension isn't yet explored. We first demonstrate an encryption framework by transforming input ciphertexts into the predesigned intensity and phase distributions within different polarization channels, and the target information can be revealed by directly synthesizing the intensity and phase information. Then, we extend to convert input ciphertexts into two orthogonal polarization channels, in which each polarization channel contains both the converted intensity and phase information. The target information can be achieved by decoding the intensity and phase information and leveraging the specific mathematical operation. Moreover, we propose a multi‐dimensional encryption architecture by introducing wavelength into the intensity and phase‐based confidential system. The robust approach for information encryption using intensity and phase of diffractive processing provides flexible potential in high‐security information transmission, data exchange, and communications.