AbstractVector quantization (VQ) technique is widely used in digital image compression. Some reversible data hiding schemes have been proposed to embed secret data (eg, password, timestamp, and watermark) into a VQ index table. In real application scenarios, users usually train codebooks on their own data sets instead of using the public codebooks directly. Therefore, different from the conventional schemes, in this paper we propose a novel reversible data hiding scheme for VQ codebooks instead of VQ index tables. All permutation operation is adopted to a sorted codebook to find usable codewords and embeds secret data. Experimental results demonstrate that the embedding rates of our proposed scheme for different‐sized codebooks are appreciable. The embedded codebook can also be directly applied to decompress images.