6 ms·

When it comes to crypto most people who read HN aren't close to being qualified to read the code and spot issues. The things the NSA is doing isn't putting in

When it comes to crypto most people who read HN aren't close to being qualified to read the code and spot issues. The things the NSA is doing isn't putting in "IF password = 'Joshua" then LetMeIn()", it's subtle flaws in random number generators. The reality is that open source may be our best defence but it's still a pretty poor defence because of the complexity of what's been compromised. The more I think about it the more I think the actual best thing you can do is to pick obscure technologies that were probably too small to warrant NSA attention.

cperciva 13y ago
I've dealt with a lot of bugs in FreeBSD, and most of them were found by non-security-experts. The people finding said bugs didn't necessarily know enough to be able to exploit the bugs, or even enough to be certain that there was a bug; but there were a lot of "this code looks weird" reports turning out to expose vulnerabilities. I don't think you need to be a security or cryptography expert to find vulnerabilities the NSA might try to exploit.
- Tyrannosaurs 13y ago
  While you're right that fewer issues means fewer potential exploits, all you're really doing is getting rid of weaknesses that should never have been there in the first place, but you're doing nothing to address the systematic undermining of encryption standards and hardware and software implementations of them. To use a metaphor, it's great that you're making sure your windows are shut but the locksmith may have sold you a deliberately weak front door lock.
  - bad_user 13y ago
    When speaking of encryption and "systematic undermining of encryption standards" you have to be aware of what that means. What people fear is the deliberate weakening of the algorithm such that it can be broken (implementations can be fixed, the real nightmare being subverted algorithms), where broken doesn't mean what many people think it means. Say you've got a 256 bits key, to brute force it you'd normally need to try 2^256 combinations of bits, right? Well, if you found a flaw that permitted you to brute force it in 2^200 attempts, then that's a massive improvement and you can consider the algorithm broken, but guess what - that's still exponential complexity with the issue being solvable by simply making the key bigger. And there are people working on these standards that are not on NSA's payroll and outside the US jurisdiction, people that aren't idiots, so bigger flaws than this aren't feasible. This is why, even if they've introduced subtle flaws in current standards, that doesn't mean they have the capability of breaking the encryption - e.g. it is possible that they are able to break RSA-1024 keys, but RSA-2048 is an entirely different problem. And RSA-4096 keys will likely stay unbreakable, unless a huge breakthrough happens. People give them more credit than they deserve: yes, they have cash and authority and can coerce companies and individuals and they can also plan for the long term, etc, etc, but let's be realistic about their abilities.
- a3n 13y ago
  That probably works in FreeBSD, but I can't imagine being welcomed by Linus if you report that "this code looks weird." Although it might be worth risking the anger, just so one would have a personal rant to hang on the wall, similar to how people save checks from Knuth.
  - makomk 13y ago
    He ought to welcome it, since I get the impression his approach to debugging relies partly on spotting code that looks weird. Of course whether he would is another matter.
- parennoob 13y ago
  > I don't think you need to be a security or cryptography expert to find vulnerabilities the NSA might try to exploit. Yes, but I'm assuming you have to be a medium-to-expert level programmer, generally in C, at least, correct? What I'm trying to say is that that is probably not the case for most people in the world, or even in the tech industry.
anon1385 13y ago
How many of the people talking about 'many eyes' on HN feel that they would have spotted the Debian OpenSSL bug[1]? As far as I remember it was not spotted through code review, but because people discovered identical certs in the wild. [1] https://www.schneier.com/blog/archives/2008/05/random_number_b.html https://www.schneier.com/blog/archives/2008/05/random_number...
- makomk 13y ago
  It's possible someone could've - just the diff in itself was fishy, because it commented out two lines which added entropy, yet it was obvious from the diff alone that that the OpenSSL developers only thought one of them should be disabled to keep memory checkers happy.
  - Tyrannosaurs 13y ago
    But "could have" is the theoretical defence. "Did" (or rather "didn't") is the real world test. This is the whole point - OSS offers the potential to be safer but if there aren't qualified people with the skills and expertise digging into every nook and cranny of every bit of code. Ironically it's like the security services say - they have to win every time to win, the terrorists only have to win once. The community have to spot and fix every exploit to secure the system, the NSA only have to get one in to compromise it - and these aren't like accidental flaws, they're things the NSA will know are there, know the extent of and know precisely how to take advantage of them.
  - jedbrown 13y ago
    This is an argument for vanilla distributions. The most sensible place for expert code review (a limited resource) is the upstream project. By comparison, distro modifications receive relatively little review. If the Debian patch had been mailed to openssl-dev, I doubt it would have quietly gotten "LGTM" replies and been picked up by a maintainer.
bad_user 13y ago
Open-source is orders of magnitude better, simply because you can compromise it only through systematic long-term efforts for introducing subtle flaws that are hard to detect by regular code reviews, and you can count on such flaws only for limited periods of time, not to mention that architectural decisions can always happen without the NSA being able to prevent it. E.g. if your Linux distro is up to date, the older flaws are likely to be fixed already. On the other hand, NSA can simply demand the addition of a dumb IF just like you described in any proprietary software, and nobody will ever find out about it. If Windows XP contained backdoors in 2001, it may as well contain the same backdoors now. Just because you can't get a 100% guarantee of safety, that doesn't mean that open-source isn't much safer. And yeah, people exemplify with that embarrassing SSH flaw in Debian. Well, at least you found out about it, at least you know it was fixed.
- Tyrannosaurs 13y ago
  That would be the sort of systematic long-term efforts we know the NSA have been undertaking and almost certainly continue to take? Given that the assumption that most people would (reasonably) make is that OSS is more resistant to these things, if you're the NSA isn't that exactly what you'd target? And given that we know that they've undermined standards, compromised hardware and so on, wouldn't it be naive to think that they'd not made significant efforts to infiltrate and compromise OSS projects? I certainly agree safer, I'm just really not convinced that safer crosses the line to actually being safe.
  - girvo 13y ago
    Frankly, against a State sponsored and run adversary like the NSA, with the clout of the US of A behind it... I personally don't think it's possible to cross the line is to safe. Frankly, with a lot of our hardware and software coming out of companies in the US, there are so god damn many avenues the NSA can take, I'm going to go fashion myself a tin foil hat at this point! FOSS gets us closer, but without FOSS hardware known (and can you ever know?) to be secure and free of influence... And I don't see the very act of creating that FOSS hardware to replace my devices as feasible at all. Sigh. This sucks. I knew it was bad, I use encryption, I share only what I want to share, etc. And none of it turns out to really matter. The NSA's actions have made me as a non-US citizen or resident so wary of anything coming out of that country now :( And with China having the same issues, the two big tech powerhouses... What's left? My own country (Australia) definitely would cooperate with these sorts of things too. We are effectively controlled by tech dictators at this point, and no way out in the near future. This sucks.

cperciva 13y ago

I've dealt with a lot of bugs in FreeBSD, and most of them were found by non-security-experts. The people finding said bugs didn't necessarily know enough to be able to exploit the bugs, or even enough to be certain that there was a bug; but there were a lot of "this code looks weird" reports turning out to expose vulnerabilities. I don't think you need to be a security or cryptography expert to find vulnerabilities the NSA might try to exploit.

Tyrannosaurs 13y ago
While you're right that fewer issues means fewer potential exploits, all you're really doing is getting rid of weaknesses that should never have been there in the first place, but you're doing nothing to address the systematic undermining of encryption standards and hardware and software implementations of them. To use a metaphor, it's great that you're making sure your windows are shut but the locksmith may have sold you a deliberately weak front door lock.
- bad_user 13y ago
  When speaking of encryption and "systematic undermining of encryption standards" you have to be aware of what that means. What people fear is the deliberate weakening of the algorithm such that it can be broken (implementations can be fixed, the real nightmare being subverted algorithms), where broken doesn't mean what many people think it means. Say you've got a 256 bits key, to brute force it you'd normally need to try 2^256 combinations of bits, right? Well, if you found a flaw that permitted you to brute force it in 2^200 attempts, then that's a massive improvement and you can consider the algorithm broken, but guess what - that's still exponential complexity with the issue being solvable by simply making the key bigger. And there are people working on these standards that are not on NSA's payroll and outside the US jurisdiction, people that aren't idiots, so bigger flaws than this aren't feasible. This is why, even if they've introduced subtle flaws in current standards, that doesn't mean they have the capability of breaking the encryption - e.g. it is possible that they are able to break RSA-1024 keys, but RSA-2048 is an entirely different problem. And RSA-4096 keys will likely stay unbreakable, unless a huge breakthrough happens. People give them more credit than they deserve: yes, they have cash and authority and can coerce companies and individuals and they can also plan for the long term, etc, etc, but let's be realistic about their abilities.
a3n 13y ago
That probably works in FreeBSD, but I can't imagine being welcomed by Linus if you report that "this code looks weird." Although it might be worth risking the anger, just so one would have a personal rant to hang on the wall, similar to how people save checks from Knuth.
- makomk 13y ago
  He ought to welcome it, since I get the impression his approach to debugging relies partly on spotting code that looks weird. Of course whether he would is another matter.
parennoob 13y ago
> I don't think you need to be a security or cryptography expert to find vulnerabilities the NSA might try to exploit. Yes, but I'm assuming you have to be a medium-to-expert level programmer, generally in C, at least, correct? What I'm trying to say is that that is probably not the case for most people in the world, or even in the tech industry.

anon1385 13y ago

How many of the people talking about 'many eyes' on HN feel that they would have spotted the Debian OpenSSL bug[1]? As far as I remember it was not spotted through code review, but because people discovered identical certs in the wild. [1] https://www.schneier.com/blog/archives/2008/05/random_number_b.html https://www.schneier.com/blog/archives/2008/05/random_number...

makomk 13y ago
It's possible someone could've - just the diff in itself was fishy, because it commented out two lines which added entropy, yet it was obvious from the diff alone that that the OpenSSL developers only thought one of them should be disabled to keep memory checkers happy.
- Tyrannosaurs 13y ago
  But "could have" is the theoretical defence. "Did" (or rather "didn't") is the real world test. This is the whole point - OSS offers the potential to be safer but if there aren't qualified people with the skills and expertise digging into every nook and cranny of every bit of code. Ironically it's like the security services say - they have to win every time to win, the terrorists only have to win once. The community have to spot and fix every exploit to secure the system, the NSA only have to get one in to compromise it - and these aren't like accidental flaws, they're things the NSA will know are there, know the extent of and know precisely how to take advantage of them.
- jedbrown 13y ago
  This is an argument for vanilla distributions. The most sensible place for expert code review (a limited resource) is the upstream project. By comparison, distro modifications receive relatively little review. If the Debian patch had been mailed to openssl-dev, I doubt it would have quietly gotten "LGTM" replies and been picked up by a maintainer.

bad_user 13y ago

Open-source is orders of magnitude better, simply because you can compromise it only through systematic long-term efforts for introducing subtle flaws that are hard to detect by regular code reviews, and you can count on such flaws only for limited periods of time, not to mention that architectural decisions can always happen without the NSA being able to prevent it. E.g. if your Linux distro is up to date, the older flaws are likely to be fixed already. On the other hand, NSA can simply demand the addition of a dumb IF just like you described in any proprietary software, and nobody will ever find out about it. If Windows XP contained backdoors in 2001, it may as well contain the same backdoors now. Just because you can't get a 100% guarantee of safety, that doesn't mean that open-source isn't much safer. And yeah, people exemplify with that embarrassing SSH flaw in Debian. Well, at least you found out about it, at least you know it was fixed.

Tyrannosaurs 13y ago
That would be the sort of systematic long-term efforts we know the NSA have been undertaking and almost certainly continue to take? Given that the assumption that most people would (reasonably) make is that OSS is more resistant to these things, if you're the NSA isn't that exactly what you'd target? And given that we know that they've undermined standards, compromised hardware and so on, wouldn't it be naive to think that they'd not made significant efforts to infiltrate and compromise OSS projects? I certainly agree safer, I'm just really not convinced that safer crosses the line to actually being safe.
- girvo 13y ago
  Frankly, against a State sponsored and run adversary like the NSA, with the clout of the US of A behind it... I personally don't think it's possible to cross the line is to safe. Frankly, with a lot of our hardware and software coming out of companies in the US, there are so god damn many avenues the NSA can take, I'm going to go fashion myself a tin foil hat at this point! FOSS gets us closer, but without FOSS hardware known (and can you ever know?) to be secure and free of influence... And I don't see the very act of creating that FOSS hardware to replace my devices as feasible at all. Sigh. This sucks. I knew it was bad, I use encryption, I share only what I want to share, etc. And none of it turns out to really matter. The NSA's actions have made me as a non-US citizen or resident so wary of anything coming out of that country now :( And with China having the same issues, the two big tech powerhouses... What's left? My own country (Australia) definitely would cooperate with these sorts of things too. We are effectively controlled by tech dictators at this point, and no way out in the near future. This sucks.