https://github.com/Anarios/return-youtube-dislike/blob/5c738...
https://github.com/Anarios/return-youtube-dislike/blob/5c738...
It could work similar to how haveibeenpwned works - send a prefix of the video ID and respond with a list of all matching IDs with that prefix. The server only knows the list not the actual video. The client can pull the correct ID out of the list.
(this was how HIBP worked before at least, IIRC)
Hell, a YT ID is 11 characters in a base64 character set. While a lot of possibilities, I do think the entire domain can be precomputed for some amount of costs.
Either way, I'm fine with this type of "leak" of data, as it's fundamental to an open web and can't be easily solved without cryptographic/hashing hoops. What's next, you want anonymity from the server that you're requesting content from, really?