While using the millisecond length of things would be clever and would catch which copy it originated from while it remained secret that was how you were doing

While using the millisecond length of things would be clever and would catch which copy it originated from while it remained secret that was how you were doing it. As soon as it was known it is not much harder for the person ripping the copy to shorten all the scenes by a random number of milliseconds than it is for them to blur the watermark. I mean this could be a good addition to the current tech used in something like Cinavia, which is the most robust system I have seen and if not already doing it could easily identify individual sources. But the inaudible audio and steganography used in that system just seem much more of a pain to circumvent than using lengths of scenes as a serial number of sorts.