4 ms·

Everything can be backdoored, but what about costs? If the interconnection between the baseband and SoC is restricted in a way baseband can't just do DMA reque

Everything can be backdoored, but what about costs? If the interconnection between the baseband and SoC is restricted in a way baseband can't just do DMA requests and mess with the system, the compromised baseband has to talk to the the compromised SoC to compromise the OS. NSA have to either develop a specific backdoor for a specific device (or, better say, schematically similar devices group) and hook into their supply chain (hmmmm...), or develop a quite cost-adding generic backdoor system. They surely have resources to design anything and even beyond that, but added cost to the production just can't be easy to conceal. There must be some sane limit to NSA's possible omnipotence. And then there are reverse engineers who love to peek what's done in silicon. NSA has to shut up not only original part vendor (easy for them, sure), but a bunch of engineers around the world, stealing designs for the chips so they'd make a clone. You know, some countries are famous for that stuff. Add: as for base stations - they're outside the phone. We don't consider plaintext data outside to be secure. If you're about the voice calls - E2E-encrypted VoIP to the rescue.

dogma1138 11y ago
The cost is irrelevant, the NSA doesn't calculate how much it would cost them to tap Bob's phone, but how much it would hypothetically cost them if they couldn't. Considering from the NSA exploit catalog we've seen that they target very specific and niche devices I don't think they care about scale when it's not possible. Yes the base band doesn't have DMA since it's connected over USB but it still wouldn't trust it if my life would depend on it in even the slightest of ways. For all we know the baseband and all basebands are compromised to the point where the NSA can tap into them DMA or not, there might be some undocumented remote debugging interface that opens a serial connection to the baseband over cellular, there could be a 1000's other things. And while USB does not support DMA it still doesn't mean it's safe by any means, attack over USB can still happen. As for NSA's omnipotence, yes there is a limit to it, but that limit won't be reached by a group of engineers building a phone with commercial off-the-shelf hardware and open source software. If the NSA's reach could be that easy to defeat than they would be very very bad at their jobs, in fact a simple commercial device like this that some how even remotely limits their ability to task their targets would be a reason to shake the NSA up completely and light a fire under their asses because they've been sleeping on the job. But as we all know they aren't sleeping on their job, in fact they act like a bunch of hyper intelligent teenagers on adderall, the NSA and their counter parts shown us for the past 70 years that no system, no network, no form of communication device is safe from them, that's their job, doesn't mean that they should have the mandate to do it all the time, but if they can't they aren't fulfilling their role.
- drdaeman 11y ago
  The cost of backdooring is put not on NSA but on hardware vendors. They have to add extra silicon and that stuff costs. And there is a huge market of reverse-engineered mostly-compatible clones, too. My idea was that I'm practically sure that if I'd take a small-enough MCU or FPGA it'd be NSA-free. Just because putting a backdoor there (and that backdoor has to be quite smart and listen for signals on a lot of pins, while being discrete about that) would seriously increase complexity and cost of the device. And that would be noticeable. Just don't believe this would went unnoticed for any long. As for SoCs - yes, they're complex enough and their interconnections are quite standardized. So, you're probably right.
  - dogma1138 11y ago
    The NSA doesn't have to tamper with hardware to backdoor it, they can just as easily find a hardware/microcode bug. And as far tampering with the actual silicon goes, well if the device is complex enough to require substantial logic then you'll probably have room to plant a bug, a simple device won't need a complicated bug to begin with. And silicon tampering doesn't require you to implement an entire bug in the silicon it self it can be as simple as intentionally added flaw that causes an error or an errata when say exposed to certain radio frequency which in conjunction with other external or internal attacks might lead to an effective backdoor. We can build fractal antenna's on a tiny scale these days http://nextbigfuture.com/2013/06/nanoscale-etching-of-3d-fractal.html http://nextbigfuture.com/2013/06/nanoscale-etching-of-3d-fra... Incorporating something like that into silicon won't be that expensive, and all that it needs to do is maybe short 2 pins that put the chip into debug mode and GL discovering that in a postmortem.

dogma1138 11y ago

The cost is irrelevant, the NSA doesn't calculate how much it would cost them to tap Bob's phone, but how much it would hypothetically cost them if they couldn't. Considering from the NSA exploit catalog we've seen that they target very specific and niche devices I don't think they care about scale when it's not possible. Yes the base band doesn't have DMA since it's connected over USB but it still wouldn't trust it if my life would depend on it in even the slightest of ways. For all we know the baseband and all basebands are compromised to the point where the NSA can tap into them DMA or not, there might be some undocumented remote debugging interface that opens a serial connection to the baseband over cellular, there could be a 1000's other things. And while USB does not support DMA it still doesn't mean it's safe by any means, attack over USB can still happen. As for NSA's omnipotence, yes there is a limit to it, but that limit won't be reached by a group of engineers building a phone with commercial off-the-shelf hardware and open source software. If the NSA's reach could be that easy to defeat than they would be very very bad at their jobs, in fact a simple commercial device like this that some how even remotely limits their ability to task their targets would be a reason to shake the NSA up completely and light a fire under their asses because they've been sleeping on the job. But as we all know they aren't sleeping on their job, in fact they act like a bunch of hyper intelligent teenagers on adderall, the NSA and their counter parts shown us for the past 70 years that no system, no network, no form of communication device is safe from them, that's their job, doesn't mean that they should have the mandate to do it all the time, but if they can't they aren't fulfilling their role.

drdaeman 11y ago
The cost of backdooring is put not on NSA but on hardware vendors. They have to add extra silicon and that stuff costs. And there is a huge market of reverse-engineered mostly-compatible clones, too. My idea was that I'm practically sure that if I'd take a small-enough MCU or FPGA it'd be NSA-free. Just because putting a backdoor there (and that backdoor has to be quite smart and listen for signals on a lot of pins, while being discrete about that) would seriously increase complexity and cost of the device. And that would be noticeable. Just don't believe this would went unnoticed for any long. As for SoCs - yes, they're complex enough and their interconnections are quite standardized. So, you're probably right.
- dogma1138 11y ago
  The NSA doesn't have to tamper with hardware to backdoor it, they can just as easily find a hardware/microcode bug. And as far tampering with the actual silicon goes, well if the device is complex enough to require substantial logic then you'll probably have room to plant a bug, a simple device won't need a complicated bug to begin with. And silicon tampering doesn't require you to implement an entire bug in the silicon it self it can be as simple as intentionally added flaw that causes an error or an errata when say exposed to certain radio frequency which in conjunction with other external or internal attacks might lead to an effective backdoor. We can build fractal antenna's on a tiny scale these days http://nextbigfuture.com/2013/06/nanoscale-etching-of-3d-fractal.html http://nextbigfuture.com/2013/06/nanoscale-etching-of-3d-fra... Incorporating something like that into silicon won't be that expensive, and all that it needs to do is maybe short 2 pins that put the chip into debug mode and GL discovering that in a postmortem.