At some point, a combination of manual/explicit reputation tagging and a web-of-trust system might become a plausible defence against a lot of these user-hostile connections.

If your corporate site, where you advertise your new high-end TVs and laptops, starts triggering warnings in all your visitors' browsers that it might be associated with malware and falling off all the SERPs for similar reasons, you're going to stop loading it with junk pretty quickly (and its trust score will improve pretty quickly as a direct result). Likewise, if your smart device tries to phone home and home has acquired negative rep, maybe that connection gets blocked automatically at the firewall.

We'd need a system that was guided by interested/aware participants who are unlikely to be successfully gamed and that mostly "just worked" for average users, but we've managed to build those in other contexts before so it doesn't seem completely out of the question.