A foundation for watermarking in compressed domain

In this letter, we propose a fragile digital watermarking algorithm that is directly applicable to compressed bit streams and is mathematically lossless with practically no increase in file size. The algorithm exploits the inefficiency of the code space of popular compression standards by recognizing that not all code words appear in the bit stream. By mapping the code words to the outside of the used code space, we can embed the watermark into the stream as forced bit errors. The decoder recognizes such code words, recovers the watermark from them, and restores the media to its original state. As proof of the concept, the algorithm is then applied to the JPEG bit stream.