Acoustic waves are promising for information encryption in electromagnetic shielding environments,or underwater.However,current encryption methods using acoustic waves are prone to information leakage during the decryption process due to their limited information capacity.Herein,we propose and experimentally demonstrate a novel acoustic holographic encryption framework based on cascaded acoustic holography,enabling encrypted information to be spatially split into two acoustic holographic plates(AHPs).To achieve this,we introduce a physics-enhanced cascaded acoustic hologram deep neural network method that inversely optimizes the phase offset distributions of AHPs.Both numerical and experimental results show that each AHP serves as a spatially separable secret key,carrying a portion of the encrypted information as a unique holographic image.Notably,the complete encrypted image,which differs from the individual holographic images generated by each AHP,is only revealed when both AHPs are appropriately cascaded along the diffraction path.This significantly enhances both information capacity and security.Moreover,we present an enhanced acoustic holographic encryption scheme that allows for the encryption of two holographic images using just three AHPs.A distinct example of underwater communication based on the proposed cascaded acoustic holographic encryption framework is further demonstrated,highlighting its capacity for high-capacity and enhanced-security parallel transmission of multiple messages to multiple receivers.With the advantages of high security,high scalability,and high fidelity,our cascaded acoustic holographic encryption framework has promising applications in fields such as acoustic encryption and underwater communication.