Physical access == compromise even for devices that are as simple as a hollow metal box.
http://en.wikipedia.org/wiki/Safe
Security ratings for those devices are measured in time. Basically, if you lose possession, it's just a matter of time. Digital security is both easier and harder, because all you're protecting there is information. If you wish for the information to be destroyed on tampering, then your job may be easier.
The only way for there to be hardening when physical access is lost is to have some form of layered defense in depth, the aim being giving the user enough time to send a command to wipe the device.
If you really want to do the analogy thing, the DMA vulnerability would be the equivalent of a safe with a door where no key is needed in the back. It would not be a very good safe.
Just sayin'.
More like a safe where the "master key" was leaked and wasn't disabled in the models that were sold.
Welcome to the world of security, I guess.
Only if it's unattended. You can't break a safe, without looking suspicious. You can't disassemble a PC, and take out its hard drive, and not attract a bit of attention.
Being able to root a system by attaching a dongle is a whole different story. It's like auto-play on USB all over again.
Yes, it goes both for a safe and a computer, so you're reinforcing my point about the equivalency of their security. Safes wouldn't be secure at all without the vigilance of bank employees, etc.
With a Firewire device to DMA the password, all you need is to hook up the device long enough to copy all of active memory. Certainly something that could happen at a hackerspace or at a conference. For the James Bond set, invent a device that you can set about 4 feet (1.3 meters) away on the table, and after it's done copying memory, the firewire cord unhooks itself and retracts back into the device.
Do you leave your PC or in standby when traveling? When you leave your desk at work?
More importantly, do you think that end users would expect password protection to work? Even when their PC is on?
There's plenty of scenarios where a PC may end up being in another persons control while powered on. This is a relevant threat scenario. Deal with it.
I'll give you that, but there are too many kinds of physical access attacks to even consider aiming at solving the entire class of attacks.
> Physical access should not equal compromise.
And I want a million dollars. Guess which one's more likely.
If someone is trusted to get access to the actual device then it's pretty much game over. Do you check the keyboard cables for key-capture hardware? Do you check for all the other nefarious devices? Do you check your OS has not been tampered with?
There's no good reason why this vulnerability still exists after 10 years except a failed design, laziness on the part of OS developers and that security professionals in general meet the problem with the above statement that "physical access equals compromise".
I think end users deserve (and expect) secure devices, even when physical access is lost. I realize that it's harder to protect a physical device, but it's not impossible.