Someone needs to put a network sniffer on a bunch of different TVs and see exactly what traffic they're generating and how much. Sending 120 4k images per minute (or worse, 6000 for the LG) on a residential WiFi connection should be pretty obvious, and it should be fairly easy to block too.
Yes. A sequence of consecutive hashes would match to a TV show or movie identifier in a local database; and that identifier is all that needs to be uploaded until there's a new TV show or movie identified.
Use one of those perceptual hashing algorithms that work across noise and resolutions. Then you just need a tiny fingerprint sent per scan that can be aggregated across consumers.
What if any traffic is encrypted?
Taking snapshots is not sending them in the meaning that software can delete similarly looking images by their own definition of similarness.
And BTW that kind of testing is more needed for some industry critical equipment - processors with ME, routers, etc. TV is a brainrot thing for dumbs, nobody needs a TV which doesn't spy you.
Definitely not users (useds?) of proprietary software. Use monitors for privacy they do not have any urge to phone home.
I agree that the testing should be ubiquitous but if the device is crooked it is no sense in trying to figure up how exactly crooked it is. Maybe it uses steganography, maybe it is able to detect MiTM devices, some TVsets cease to keep working if the internal memory is 100% filled with pictures of what are you watching so no extra knowledge about the crooked device is useful if you have already agree to have Pavlik Morozov with microphone and videocam right in your bedroom.
