21–30 of 41 posts
This is a mostly useless article. Every computer with USB ports has this problem. If you want to use it as, you know, a computer, you're probably gonna need to plug some stuff into it at one time. You could make the argument that now that it has one port, it's easier to secure because you only have to secure one side of the computer instead of two.
USB Type C does not support DMA as the article states. I think there might be some confusion on the attack vector here. BadUSB requires cooperation from the OS/BIOS/a driver.
We went through this with "autorun" on CDs and DVDs. For years, Windows would run anything that looked executable if you put it in the obvious place. That didn't end well. Those defaults were finally changed.
What's needed is to turn off automatic hot plugging for USB devices. The era when you can plug in a USB device and have it go live without user interaction is over. We're going to need clear OS dialogs - "The USB device you just plugged in claims to be a keyboard - did you just plug in a keyboard?" You should't be able to boot from a USB device without doing something to enter a maintenance mode.
There's nothing special about USB Type C here. It's just that the asymmetry of USB is being dealt with.
Asymmetrical USB has created an amusing hierarchy. Desktop computers were masters. Phones were slaves. Then came tablets. Are they masters or slaves?
Earlier quoted context omitted.
Are you kidding? It means that if you go to a coffes hop and plug your laptop, you can become compromised. Sure, random people in a starbucks won't, but it's a boon for industrial espionage!
Just wondering, how does that mean that? The way I understand things that is only possible if you use someone else's compromised power adapter. I've never seen a coffee shop that supplied power adapters, it's always just a bunch of power outlets. I think your safe for now.
This is written like it is a novel situation. Smart phones have had this problem almost since inception. Would be nice to see a hardware switch to deactivate the pins that aren't needed for power.
That switch sounds like a tech' support nightmare. Even WiFi switches are a massive PITA that are almost more trouble than they're worth. I'd suggest that if people want safe USB that they just buy a USB "condom" (i.e. an adapter that goes between the power and port, and disconnects the data wires). Although does USB-C use any of the data pins to determine if it is safe to transmit power? I just ask because Apple's U…
Power Delivery 2.0 signals via Vbus (the 5V line) and additionally via CC when using the Type C connector. PD2 is specified for Type A, Type B and Type C connectors.
This is written like it is a novel situation. Smart phones have had this problem almost since inception. Would be nice to see a hardware switch to deactivate the pins that aren't needed for power.
Edit: This will presumably disable any usage of the non-power lines.
^^ Oh yeah, that would do it. Damn Apple for disabling the tape-over fix!
The article (especially the headline) implies that the new Macbook is somehow more at risk because of its reliance on one USB port. But if the root of the problem is USB more generally, wouldn't the only "secure" design preclude the inclusion of any USB ports whatsoever? My issue with the article is that by singling out the Macbook (due to its single port) they are not actually contributing any useful information bac…
The article also ignores that many of these USB issues are actually solvable in software (both uEFI and in the OS). The only USB issue which REALLY isn't solvable is the fact that USB devices can simulate a USB hub and then attach fictional human interface devices (i.e. keyboards and mice) which they can then use to take over the computer and or otherwise cause problems.
USB Type C does not support DMA as the article states. I think there might be some confusion on the attack vector here. BadUSB requires cooperation from the OS/BIOS/a driver.
Right. A USB device can't do anything unless the OS lets it. We went through this with "autorun" on CDs and DVDs. For years, Windows would run anything that looked executable if you put it in the obvious place. That didn't end well. Those defaults were finally changed. What's needed is to turn off automatic hot plugging for USB devices. The era when you can plug in a USB device and have it go live without user intera…
The problem with that is that you inevitably run into this dialog with nothing else attached to the machine. How are you going to confirm the dialog?
Better to focus on isolating the keyboard driver from the rest of the OS, so the only thing it should be able to do is read in input from the keyboard and report it back to the OS.
USB Type C does not support DMA as the article states. I think there might be some confusion on the attack vector here. BadUSB requires cooperation from the OS/BIOS/a driver.
Right. A USB device can't do anything unless the OS lets it. We went through this with "autorun" on CDs and DVDs. For years, Windows would run anything that looked executable if you put it in the obvious place. That didn't end well. Those defaults were finally changed. What's needed is to turn off automatic hot plugging for USB devices. The era when you can plug in a USB device and have it go live without user intera…
Of course there is the chicken-and-egg issue of allowing the first keyboard you plug into a machine which does not have any built-in input device.
Earlier quoted context omitted.
Right. A USB device can't do anything unless the OS lets it. We went through this with "autorun" on CDs and DVDs. For years, Windows would run anything that looked executable if you put it in the obvious place. That didn't end well. Those defaults were finally changed. What's needed is to turn off automatic hot plugging for USB devices. The era when you can plug in a USB device and have it go live without user intera…
> The USB device you just plugged in claims to be a keyboard - did you just plug in a keyboard? The problem with that is that you inevitably run into this dialog with nothing else attached to the machine. How are you going to confirm the dialog? Better to focus on isolating the keyboard driver from the rest of the OS, so the only thing it should be able to do is read in input from the keyboard and report it back to t…
That's only a serious issue for a keyboard, and for this specific situation the keyboard could be segregated (globally inactive) and require a specific sequence of action to be activated.
> Better to focus on isolating the keyboard driver from the rest of the OS, so the only thing it should be able to do is read in input from the keyboard and report it back to the OS.
The virtual keyboard can send whatever keypresses (and thus indirectly commands) it wants to the system.