Maybe a simple way to get around server side ad injection would be to have a browser extension that hashes every, say, 1 in 10 frames of the video stream, then compares it to a central database of hashes for ads and detects if it should be skipped. You would have to store different hashes for each ad, and for each resolution and codec, and routinely crawl YT to collect every ad for every region, but if it comes to that, clearly the motivation is there for people to try it.
A potential optimization might be using perceptual hashing instead of the normal kind.
[URL unfurl="true"]https://en.wikipedia.org/wiki/Perceptual_hashing[/URL]
