One chestnut from my history in lottery game development:
While our security staff was incredibly tight and did a generally good job, oftentimes levels of paranoia were off the charts.
Once they went around hot gluing shut all of the “unnecessary” USB ports in our PCs under the premise of mitigating data theft via thumb drive, while ignoring that we were all Internet-connected and VPNs are a thing, also that every machine had a RW optical drive.
Pretty easy to make a hub device that you can plug the keyboard into and make it transparent to the user. Could even build in a keylogger to capture direct from the keyboard. The attacker would likely need physical access for that, so it wouldn’t be as convenient as the thumb drive in the parking lot attack vector, but unless you’re using PS/2 peripherals (or gluing those USB devices in too somehow), there’s still a fairly open attack vector there, even if you are disabling unused ports in BIOS.
Yep you’re right, but at least that adds another layer of complexity to their attack. A lot of security controls are at least somewhat situational, and most non-draconian companies have a process to put further mitigations around those exceptions either from increased monitoring or adding additional supplemental controls.
There’s no such thing at perfect security, just better risk mitigation. Slipping in a usb hub between the computer and keyboard while someone isn’t looking is a bit trickier then just plugging in a usb stick. If you disable unused usbs in the bios, instead of trying to do silly stuff like glue them shut, then the attacker has at least been temporarily thwarted if they slot it into a dead port. Aside from the high traffic areas, disabling ALL usb ports in places like datacenters and especially colocated datacenters, can thwart the attack outright as well.
Really from looking through this thread a lot of people seem to be under the misconception that security that isn’t perfect is pointless. It’s like claiming that locking your doors is pointless because lockpicks exists. The point isn’t to keep a sophisticated attack at bay, but rather to keep script kiddies and drive-by attacks from hitting your network. To defend against sophisticated attacks you really have to go a bit crazy, and even then very small slip ups can be disastrous. Ask Microsoft about their root cert getting leaked via a core dump!
I fully acknowledge that many people also work for places with dumbass security controls. Gluing usbs is WAYYYY up there on that list in my opinion. It also looks like a lot of people work at places that have really shitty security teams that haven’t quite figured out that controls are situational and require more thought then, “see checkbox, execute checkbox.”
If it’s a secure enough environment, I imagine that there will be monitoring on the device, and the moment a hub shows up that’s not supposed to be there, or any other USB device tree that doesn’t match the approved list, , alarm bells ought to go off. If it’s valuable enough; the attack would be to use a passive device picking up leaky signals on the wire, or even hidden camera watching screen/keyboard.