[comex]

> The takeaway is that your children may grow up in a world where it's impossible to guarantee the government can't get into your computer if it really wanted to.

This is impossible to guarantee today. Certainly if you run the zero-day magnets known as browsers, and even if not, there is always some possibility of physical intrusion.

> More and more network adapters seem to have DMA access to your computer.

With an IOMMU (VT-d or equivalent on other platforms), it should be possible to protect against malicious DMA from any source.

Also, not all phones have basebands with DMA access to main memory. I think iPhones do not, though I am not sure, and some older iPhones have been attacked by turning on "auto answer", demonstrating direct access to the microphone.

[devconsole]

Unfortunately, projects such as DROPOUTJEEP confirm that the iPhone isn't to be trusted.

This is impossible to guarantee today. Certainly if you run the zero-day magnets known as browsers, and even if not, there is always some possibility of physical intrusion.

Today you can use OS's such as Tails to prevent most exploits from embedding themselves into your computer. This is what Snowden used, for example. But if hardware becomes compromised, Tails will offer much less protection.

Here's an interesting section of the article:

The department must describe the computer it wants to target with as much detail as possible. For example, an investigator may be covertly communicating with a suspected child molester and know an IP address, and then obtain a warrant to use malware to find the actual location. In the case of botnets, malware might be used to try to free the compromised computers from a criminal’s control.

Imagine if child molestors begin using Tails. The government response may be to try to set up some kind of "Tails dragnet" via compromised network interfaces. It should be possible for a network adapter to detect that Tails is running. At that point, since it has DMA access, and since few people use Tails at any given time, it should be possible to instuct a network adapter to search through a computer's memory for evidence of activities that the government doesn't like. Since Tails offers strong anonymity protection, there's no way to describe a computer "as specifically as possible" other than to say "it's running Tails while watching child porn."

The unfortunate conclusion is that in the future, someone like Snowden might immediately be caught. "If someone is using a strong anonymity tool and GPG to hide their conversation, we should probably configure their network card to monitor their activity."

Once hardware begins to turn against you, there seems to be nothing anyone can do to protect themselves. Encryption doesn't work against an adversary that has access to your computer's memory.

[cam_l]

"strong anonymity tool and GPG"

just thinking, the problem, it seems, is that end-to-end encryption is not really end-to-end. the user is the endpoint, not the computer.

from a ux point of view, a dongle between screen / keyboard and computer for an encryption overlay could be a way to unambiguously protect information - so information never exists decrypted in a machine.. just the screen.

user input-output accessories are much more technologically static than software / hardware, so an open-source hardware solution may be possible?

[cinquemb]

>The unfortunate conclusion is that in the future, someone like Snowden might immediately be caught.

I think that is too naive. Snowden types don't assume they won't be caught, they probably assume that it is only a matter of time until they are caught, and play the cards they have in such a way that you make it really hard to send your garden variety cia/dia/spec ops/defense contractors out on a pick up operation not only only from a feasibility standpoint, but from a geopolitical stand point (e.g. What will Beijing's/Moscow's/D.C.'s response be if we run such an operation in their front yard? What precedents might we be setting?).

Also to note that offensive/defensive technical capabilities aren't as asymmetric as they appear for all possible targets of nation states, some yes, but probably not as much to those with the technical knowledge who can create/use such and derivative systems which might very well be other nation states (or appearing to originate from such).

[devconsole]

Also to note that offensive/defensive technical capabilities aren't as asymmetric as they appear for all possible targets of nation states, some yes, but probably not as much to those with the technical knowledge who can create/use such and derivative systems.

If you concede that your computer has a chip with DMA access which can be used by the government, then you must concede that the same chip can monitor you for activity that triggers active surveillance. For example, I think Tails is going to force governemnts into monitoring at least which operating system you're using. There's no way to target a specific Tails user, so the only recourse is for the government to do dragnet surveillance of everyone using Tails, or ignore the activities of those using Tails. Since the latter seems politically untenable, the former is becoming more likely with time. When the government can passively check whether your activity is fitting the pattern of some kind of criminal activity, the situation is about as asymmetric as I can imagine. Is there really any technical knowledge that could protect you?

[cinquemb]

>If you concede that your computer has a chip with DMA access which can be used by the government, then you must concede that the same chip can monitor you for activity that triggers active surveillance.

Whats DMA access? Direct Memory Access access?

That aside, I'm not willing to concede that across every computer than has been/can be built and be exploited by a government out of the box remotely [because dragnet] (most of them, I will concede probably can though, and conversely anyone technical enough can probably exploit many systems in the same way for their own means[don't trust your spouse/freinds/employees?, bug them with remote backups of data to analyze in real-time, hell, companies do such things now as-a-Service]). But continuing on with your conclusion of a dragnet (which is more or less present today), access isn't really the problem, but you have a signal and a noise problem, wherein you will have false positives and false negatives. Text book example of the mal-possibilities is the NSA providing data which led to the targeting phones in the ME, which drone strikes we're initiated and hit innocent civilians[0]. Just wait when we're at the point when this is happening within a countries national borders by domestic agencies, one day, someone is going to be taken out that wasn't meant to be taken out. Can't ignore the false neg/positives forever, though governments seem to try very hard to do so. I think corporations are more forthright about the extent the data they collect is able to be used because if you knowingly contract/ utilize bogus data for certain applications, someone else will eat your lunch eventually.

>activity is fitting the pattern of some kind of criminal activity

From a predatory-prey/evolutionary standpoint, criminal activity is always evolving (typically every living being and the systems they rely on are). Not to mention the time sensitive nature of these systems that do the analysis so if $criminal_activity is always changing and over a defined period of time, you risk that you will get no signal for those who conduct such $criminal_activity in less than the defined period of time or that by the time the analysis has been done, or any signals collected from such device will be moot (i.e. computer was destroyed, thrown away or even worse: passed along to/associated to someone else which also means any point there after associated with such systems is akin to going after a ghost within the machine).

>Is there really any technical knowledge that could protect you?

Well since the focus is on tails [but mostly on the dragnet], one can clone the sc[1] and go through it for what could possibly define one as a Tails user, replace that with something else, build their own image and voila, you just avoided being in the dragnet. The thing about dragnets is that they can only really capture the lowest common denominator, deviate only slightly from that, and the adversary will have to expand resources going for an targeted operation (any adversary, not just nation states is technically possible of doing these things and by definition not a dragnet). This is what happens today. Not in some far off distant dystopian future meant (intended or not) to invoke fear in the ignorant/lazy. Yes if one wants to avoid being in a dragnet with some of the tools they use, then one will take the steps necessary to keep such information obfuscated/opaque from the dragnet.

[0]: http://www.policymic.com/articles/16949/predator-drone-strik...

[1]: https://tails.boum.org/contribute/build/

[devconsole]

Whats DMA access? Direct Memory Access access?

This is what happens today. Not in some far off distant distopioan future meant to invoke fear in the ignorant/lazy.

Why not talk with me without the snark? This topic seems like it interests you a lot, so it seems like we have some shared ground.

one can clone the sc[1] and go through the source code for what could possibly define one as a Tails user, replace that with something else, build their own image and voila, you just avoided being in the dragnet.

This won't work because it's extremely difficult to analyze your network card and discover its behavior, and without this knowledge you'd be changing things blindly. There are far too many ways to detect an OS to change them all. Tweak-and-recompile would work if they use a naive and brittle heuristic like "look for the first 64 bytes of whatever is loaded into memory when Tails is booting up," but they wouldn't employ such a brittle heuristic in the first place because every time a new version of Tails is released, they'd need to update their entire infrastructure to look for a new pattern. Something like monitoring the network traffic for a unique "Tails signature" is more likely in this scenario; for example, how many computers start Tor immediately after a network card is connected? Detecting that condition would be a decent starting point for detecting Tails, and they'd want to combine it with some other hard-to-evade condition to cut down on false positives without introducing false negatives.

One interesting way to detect that someone is using Tails would be to notice that their system clock is set to UTC time. Most of the computers connected to the internet aren't using UTC, so UTC time plus Tor usage on startup is pretty commonly associated with anonymity OS's. That said, it seems like it might be difficult for the network card to detect whether the system clock is UTC time, but it's just an example of how difficult it is to fully conceal your usage of an anonymity tool. It's not just a matter of tweaking the source code.

This seems to prove the seriousness of this threat, though. Once you agree that it might be possible for your network card to be your adversary, there are endless ways that it can be used to defeat you. Hardware manufacturers have evidently been thinking along these lines, so why shouldn't we try to think of ways to prevent this from happening? As the BIOS exploits have shown, that dystopianic future may be closer than anyone's comfortable admitting.

EDIT: Someone went through and downvote bombed our whole converastion on both sides... I tried to correct it, but it looks like upvotes from Tor users under a certain karma threshold aren't registering, so I wasn't able to help fix it.

[cinquemb]

>Why not talk with me without the snark? This topic seems like it interests you a lot, so it seems like we have some shared ground.

>One interesting way to detect that someone is using Tails would be to notice that their system clock is set to UTC time. Most of the computers connected to the internet aren't using UTC, so something like that is pretty commonly associated with Tails. That said, it seems like it might be difficult for the network card to detect whether the system clock is UTC time, but it's just an example of how difficult it is to fully conceal your usage of an anonymity tool. It's not just a matter of tweaking the source code.

It's not out of snark (I apologize for if it sounds like it, not intentionally seeking to offend anyone), but mainly out frustration about the conversation on how everything seems to be so difficult. Difficulty to whom? Someone who cannot modify sc to a significant extent? Someone who just downloads the program and expects it to just work? Not just some random tweak, I mean going through looking at what the functions actually do, which remote connections do they rely on to connect to at various stages, how data is generated and allocated in memory, what system calls are made, etc and change it according to ones threat model so that the program one complies has the same functionality but is not recognized as the same program. Maybe that involves changing the the system time. Again, trying to target someone doing such is trying to target someone actively adapting, probably faster than it takes for the dragnet to adapt since like I said, dragnets mainly hinge on effectively going after the common denominator that of which is usually of the mind set of someone who downloads/uses a program system and expects it to just work and address all of their concerns without doing anything themselves. In the end, anyone can try all they want to cut down on the false negatives and positives, but they will still exist and that's where the "real" danger comes from for groups/orgs/gov's that go to such extents.

>Once you agree that it might be possible for your network card to be your adversary, there are endless ways that it can be used to defeat you.

If this is really in one's threat model, one is probably throwing away or using shared computers before this point… maybe from within a virtual machine on a large banks network from an exploit one used (remote, or local).

>so why shouldn't we try to think of ways to prevent this from happening?

Few people do this today for themselves, most others do not. People today seem to have come to expect that someone else needs to protect them which must have fmr cyhperpuks laughing. As far as I'm concerned, we are already living in the dystopian future, and the few who take the steps to mitigate based on their threat model do. These issues have been around for a while, and those who cared all along took steps they felt were necessary to protect themselves and still do. Maybe that involves not taking advantage of the latest skinner box of the day, again tradeoffs and threat models to consider. And those now made aware have to learn a lot to put themselves in the same shoes, if they even care enough to learn what they need to start protecting themselves and to continue to adapt to do so. Again, its not like BIOS exploits suddenly became possible because snowden profiteers told us and because all of this I don't think it really is a serious threat (any more than it already was) because your adversaries are opening themselves up at the same time. This has always been an evolving landscape. Such is the world we live in and have always had.

Edit: No worries, as I've learned over time, down-voting isn't really effective for silencing ideas/discussion since it just attracts more interest to those who want to seek such information.

[andreyf]

> projects such as DROPOUTJEEP confirm that the iPhone isn't to be trusted

That iPhones used to completely trust physically connected devices without any verification was obvious to anyone paying attention even before it was verified at Black Hat USA 2013 [1]. This was fixed in iOS 7. The evidence we have of DROPOUTJEEP says it is installed via "close access methods" [2]. I wouldn't be surprised if remote vulnerabilities exist that could be used to install it remotely, but I am aware of no public evidence that they are being exploited now.

1. http://www.zdnet.com/researchers-reveal-how-to-hack-an-iphon...

2. http://www.zerohedge.com/news/2013-12-30/how-nsa-hacks-your-...

[comex]

> Once hardware begins to turn against you, there seems to be nothing anyone can do to protect themselves. Encryption doesn't work against an adversary that has access to your computer's memory.

In the future (or today, depending on your setup), IOMMU. In the present, there is no evidence that baseband backdoors of this type actually exist (as opposed to hacks). When the adversary adds backdoors deeper in the hardware? ...well, we'll see if that is discovered someday.

To editorialize a bit, I guess it can't hurt to worry about and try to head off anticipated future threats - it's not like anticipating different threats is mutually exclusive - but still, I somehow can't shake the feeling that people's emphasis on secret backdoors unduly weights threats that are easier to romanticize over more pragmatic but more dangerous ones.

[devconsole]

The reason it's good to proactively think of future threats is because so many past concerns have proven to be true. Several months ago, no one on Hacker News really believed that BIOS backdoors were much of a threat. But today it's a well-established fact, for example.

The tools of law enforcement probably aren't going to be revealed, and they're hard to discover. Nobody knew about the zero-day exploit employed against Tor browser, for example, and there are almost certainly many more tricks like that up their sleeve. They already take steps to conceal them; parallel construction is an unfortunate reality. And since there's not much justification for a whistleblower to reveal the techniques, it's unlikely someone will come out and talk about them. We'll probably need to think along the lines of "What's technologically possible, and how is it useful to law enforcement?" It's not a good idea to wait until a weapon is used before thinking about how to react to it.

The history of communications technology and how governments have reacted to the technology is actually quite fascinating. Wiretaps used to be extremely commonplace, and since there's not too much legal protection from the government rifling through your digital life at will (at least compared to getting permission for wiretapping your phone), it seems like it's better to err on the side of caution.

It's also important to realize that even though some governments follow due process, several powerful ones don't. Also, there are other global other considerations. The US has made it pretty clear that their legal restrictions are designed to apply to US citizens, not any foreign person. You may be forced into a situation of choosing which governments you'll trust, especially when cross-nation collaboration becomes even more pervasive. Assuming that other countries adopt a similar attitude of "Our citizens are protected; other citizens are examined," then the US may simply outsource their databases of information to be examined by some other government, like any other member of the Five Eyes.

I understand your concern and skepticism though. It was a question I've often wrestled with myself.

[zaroth]

  Nobody knew about the zero-day exploit employed against Tor browser,
  for example, and there are almost certainly many more tricks like that
  up their sleeve.

I had actually been patched already upstream, so it was not really a zero-day. I'm not sure if it a patched Tor Browser Bundle had been released and people just hadn't upgraded, or if the patch hadn't made its way to the bundle yet.

[aliakbarkhan]

> This is impossible to guarantee today. Certainly if you run the zero-day magnets known as browsers, and even if not, there is always some possibility of physical intrusion.

Bingo. Even if you go all-out with security and only browse the web with a pure text web browser through Tor running on a VM that you purge after every use, use full-disk encryption with plausible deniability, fully shut down your computer and wait until the RAM is cool before leaving, inspect your computer for NSA/other implants every time before boot, tape your webcam and mic, never use your real name, and whatever else you can think of, you're just going to go nuts from all the paranoia, as well as from realizing all the myriad ways your security could still be broken (don't forget to check the keyboard for a built-in logger and look inside your case for PCI cards you don't recognize, and hope they don't have any implants that look convincingly like something you'd recognize as yours). Never mind that this isn't a very viable way to do most things most people actually use their computers for, like personal email, online banking, social networking, and so on.