All-optical data class-specific image encryption using diffractive neural networks

We present an all-optical data-class-specific transformation framework that encrypts image data using separate transformations uniquely assigned to each data class. The presented method was experimentally tested under terahertz and near-infrared wavelengths.