▎ 摘　　要

Secret sharing, as a sophisticated cryptographic technique, is widely used to protect important messages in the information era. As an important evaluation criterion, being able to provide timely warning and judge the security of the secret once eavesdropping occurs is a highly practical way to minimize losses. Here, the concept of countersurveillance holographic secret sharing based on laser-induced irreversible hologram reconfiguration of graphene oxides (GO) is demonstrated. The laser- and thermal diffusion-induced reduced graphene oxides (rGO) form two regions with different transmittances, which are defined as rGO superpixels. The rGO superpixels enable synchronous two-amplitude-state modulation of the holographic pixels, resulting in hologram reconfiguration. As such, multisecret sharing is realized by introducing a deep-learning gradient descent algorithm into hologram encoding, which broadens the functionality and information capacity. The results establish a high-security information delivery platform based on GO and pave the way for an all-optical on-chip holographic cryptography strategy.