A Cross-Modal Searchable Encryption Scheme with Result Verification

With the development of the Internet of Things (IoT), there is a rising demand for ciphertext retrieval. However, existing searchable encryption schemes mainly support single-modal retrieval, while current cross-modal searchable encryption methods often suffer from high computational overhead and lack reliable result verification. To address these problems, we propose a cross-modal searchable encryption scheme with result verification (VCMSE). First, we design a cross-modal hash extraction method that combines contrastive learning with a residual similarity matrix to generate encryption-friendly binary features with enhanced semantic consistency. Second, we designed a lightweight garbled circuit-based matching mechanism that enables efficient similarity computation in the ciphertext domain. Third, we propose a triple verification mechanism to ensure the search results from the cloud server are correct, complete, and comprehensive. Experimental results demonstrate that, compared with other cross-modal searchable encryption schemes, our method improves mean average precision (MAP) by 3.02%–16.9% on the NUS-WIDE dataset, while also reducing trapdoor generation time by 94.8%.