kernelpanic... Great plan; I think you're dead on about the user signing with private-key and the tablet having public-key; but I'm not sure why the server needs to authenticate? If the server is pushing messages signed with the right private-key (and hence, valid messages from the owner), isn't that good enough? Not that it hurts; just wondering why...
I'd be inclined to have arbitrary execution (as user user) through the daemon; that leaves you with the ability to wipe files, etc. at your own discretion. Also, you can then make the tablet (if you pre-equipped it with the right tools...) ssh out to your desktop, giving you a shell into your tablet when it's connected. The ability to manually fiddle around, diagnosing his network, firing up GPS and/or camera when desired (because intelligence >> heuristics), and so on, is highly valuable, imho. Streaming video and selecting which snapshot to take (and send through to the official server, for evidence purposes), you get the idea. (I won't even mention cracking all the machines on his home network, finding one with an old PCI modem still installed, and dialing out to your cell so you can get his phone number, or anything else like that.)
I can't be the first one who thought, upon reading "It should also try to connect aggressively to any unencrypted Internet connection it can to phone home and ignore normal WLAN connection settings also on a schedule You obviously can't put that in an official release.
But if the owner can run arbitrary code, that's their decision to add whatever level of zeal to their autoconfig they are willing to risk. But if the daemon's open-source, people will add such functionality anyway, so it seems like it should be included.