I don't think you actually need to do the keyboard thing.

You should be able to use "accessibility" features to bring up a soft keyboard (or just cut and paste from known things) to run the entire attack as a normal-appearing HID mouse, which just happens to be evil. No need to reconnect via a hub; you could do the whole thing within a mouse controller. Bonus points if you can manage to fit all the logic in the same processor/memory the thing includes by default, or something which looks like the same physical package and thus passes a non-destructive open-device analysis.

(which is kind of terrifying if you think about it; almost no one checks the provenance of "dumb" devices like mice. Keyboards sometimes get a tiny bit more scrutiny due to hardware keyloggers inside inline, but that's about it.)

The ultimate solution is a trusted supply chain and equipment which is tamper-evident, non-counterfeitable, and which can securely attest to intactness and provenance; x9.9 pinpads and hsms. But no one does that for general purpose hardware.

Except the mouse doesn't know exactly where it is on the screen, or how it's concept of CPI (counts per inch) relates to pixels on the screen.

That's a pain, but I think I could work around it by observing the human's use of the mouse for a while; there are some known actions (let's restrict to Windows), like clicking on start menu and some other stuff. By watching and calibrating, you could get a good idea what to do. You also might know a lot about the target environment; maybe they have totally standardized configurations including mouse speed, display size, etc.

The environment I'm thinking about is something like a SCADA control system for, say, a centrifuge cascade :)

Fascinating.

How hard would this part be?

> But surely you could simply test the hardware

Scanning system devices would reveal it pretty quickly if it's just an additional USB device.

Yes, if you took the time to connect each mouse to a machine and see what devices it detects, it would be readily apparent that a mouse loaded with a Glitch was doing something funny. And, as far as that goes, a really observant user might notice something weird going on when it's first connected. But most users aren't terribly observant, or at least that's the assumption.

A slightly cleverer device could get even better. If it only enumerated as the "evil" device (whatever it is) after several minutes or even hours after some amount of activity followed by inactivity, then it would be quite likely the user isn't at their PC to notice any messages from their operating system, suspicious behaviour, or other warning signs. Someone attempting to scan hundreds of mice for malicious ones would have an even worse time: Because it waits until the user appears to be using the mouse, simply plugging them into a test computer won't immediately reveal the evil. This can be taken further, including fingerprinting the host operating system, time of day, or any other number of factors.

Even that's not guaranteed. You could just have the mouse wait 1 hr before doing nasty-keyboard-mode. Or only go into nasty-keyboard-mode after certain amount of button presses etc. It's an arms race.

Sounds like http://hakshop.myshopify.com/products/usb-rubber-ducky