It sounds like it's good practice to get two copies of a document from two different sources, and to compare their hashes before publishing them. You can embed data in anything, so this would include images, audio files, PDF files, or programs. At least for Elsevier it's pretty obvious they're using a key to track you.