Hybrid Physical and Algorithmic Codesign for Information Encryption via Multiplexed Meta-Holography

In recent years, metasurface (MS)-enabled information encryption has emerged as a new paradigm that leverages the inherent multidimensional degree of freedom (DoF) of electromagnetic (EM) waves to construct highly secure and adaptable channels. In this paper, we extend this paradigm by developing an MS-based image encryption system that integrates multidimensional EM modulation with an additional algorithmic-layer cryptographic processing to further enhance security. In the proposed framework, a multifunctional MS acts as the physical key, while an autoencoder (AE) is introduced to retrieve the phase distributions required to configure the MS for multiplexed holographic channels across several orthogonal DoFs, including orbital angular momentum (OAM) modes, polarization states, and operating frequencies. Then, the target image is mapped onto these channels using user-configurable encoding rules, and the resulting data are further encrypted through a Feistel network and Base64 transformation. The encrypted data are finally packaged into a quick response (QR) code and transmitted to the receiver, who can recover the correct image only when both the hardware- and algorithmic-layer encryption rules are fully known. Therefore, the proposed hybrid encryption strategy offers a feasible pathway toward more versatile encrypted imaging systems and has the potential to accommodate emerging requirements in secure information exchange.