Nowadays, data-hiding techniques for images generally work by embedding the secret data in either the spatial domain or the frequency domain. By contrast, little attention has been drawn to data hiding in the compression domain. However, in general, digital images are often compressed because of the limited transmission bandwidth and storage space. If the secret information can be embedded into compressed files, then it is not likely to arouse the attention of attackers. In this paper, we shall propose a VQ-based data-hiding scheme using principle component analysis. The proposed scheme combines two techniques: vector quantization (VQ) and principle component analysis (PCA). Judging from our experimental results, the new scheme is capable of providing better image quality and embedding payload than the traditional VQ-based method.