4 ms·
When Engineers Lie
- PaulHoule 15y agohe makes a very good point. the U.S. government relaxed it's restrictions on key lengths after it came to the conclusion that vendors in general would not get key management right
- tptacek 15y agoIn the time period during which the government cared about strong cryptography, its use was almost unheard of in off-the-shelf products. By the time SSHv2 was released, shipping strong cryptography internationally was reduced to a simple bureaucratic formality. This notion that a malignant intelligence in the federal government is plotting ways to keep IT products insecure is a product of the narrative fallacy. The government may think that many steps ahead when it comes to China (or, as recent events indicate, maybe not), but when it comes to civilian information security, it is entirely reactive.
- cube13 15y agoI think the government also realized that for the majority of communications, truly strong cryptography is pretty useless. There is always a point where the information is basically worthless to keep secret. For example, if the president(or other high ranking official) is making a "secret" visit, that information only needs to be kept encrypted for fairly short amount of time(maybe a week or so), which means that a less secure algorithm that takes a few weeks to crack is enough to encrypt the information. That way, even if an enemy gets the information and cracks it, it's useless when they finally read it. For things that may need to stay secret for longer periods of time, like intelligence reports with the names of informants or long-term defense strategy plans, a stronger encryption method would be needed.
- tptacek 15y agoI have no insight into the NSA's capabilities; I've heard murblings that their practical edge is in knowing classes of implementation flaws that we don't, and in having decades of experience weaponizing those flaws... but for all we know, they have some algebraic structure attack on AES that makes it trivial to break. However, bear in mind that there's a real dollars-and-cents value to obscurity in cryptography when you're the NSA. When you're the NSA, you know your custom undocumented algorithm isn't trivially breakable (unlike virtually all bespoke commercial ciphers, which end up getting solved by SAT solvers as an exercise). Forcing your adversary to cryptanalyze an entirely new cipher imposes real and severe costs on them. That alone is a huge motivation to keep their own capabilities secret, and to not use standard algorithms to protect their own secrets --- even if, for all they know, those algorithms are sound.
- eli 15y agoHe totally lost me with the part about IPv6. What does any of that have to do with security or government eavesdropping? I read the linked article and it just says that the NSA recommends you disable IPv6 if you're not using it. Which, well... duh.
- 83457 15y agoMaybe it has to do with devices and computers being directly addressable?
- eli 15y agoAnd that makes them more secure somehow? I like Cringely, but sometimes I really don't think he has a firm grasp on what he's talking about.
- iwwr 15y agoYou can no longer rely on NAT to provide a default firewall, can you?
- tptacek 15y agoIf you enjoyed the security of NAT in IPv4, you're free to continue using it in IPv6.
- tptacek 15y agoThis is nonsensical from nose to tail. Encryption capabilities that are (on paper at least) unbreakable by any agency† are practically de rigueur in all modern applications. Where those applications are getting broken, it is not because the government is weakening algorithms. Nobody is breaking 128 bit AES, let alone 256 bit (or Cringely's fictional 4096 bit version). They're breaking applications and protocols. Similarly, IPv6 has practically nothing to do with network security. If anything, by reestablishing a static addressing regime, it makes the Internet easier to monitor. The devices that the government uses to monitor backbone connections aren't limited by protocols; they're limited by speeds and feeds. If the government has policies advising operators to turn off IPv6 on Cisco gear, it's because IPv6 has (at least until recently) made devices less stable and less secure. What worries me here is that Cringely, who has been irrelevant almost since the moment he lost his gig at PBS, is feeling revitalized by these most recent infosec stories. He is clearly ill-informed, but has enough of an audience that his name puts him on the front page of HN. People seem to feel this way about Gruber, but there is at least a point to Gruber (apart from the fact that he writes circles around Cringely): he is a passionate advocate for a perspective that is at least valid, even when we disagree with it. What's Cringely an advocate for? From what I can tell, it's "taking crumbs of information from anonymous sources and writing entire speculative stories about them". Just look at this one! "The government is suppressing strong cryptography which is why we don't have IPv6, [my sources tell me]". † Using "unbreakable" for "the disclosure of information that could only have been obtained by breaking it, thereby disclosing our capability of breaking it, would be a calamity for national security, diplomacy, and the economy".
- irishcoffee 15y agoYeah, I was pretty shocked this his the front page on HN after reading the article. Your comment gives me hope that the greater part of the NH community might agree with you. This guy has literally no idea what he is talking about. It took all I had to not comment on his article.
- akkartik 15y agoThe nose at least seems to have a point. There is no claim that 'the government is weakening algorithms'. The claim is that the government is weakening implementations -- 'applications and protocols' as you put it. He's speculating that proprietary vendors are hiding details about the break-ins because the details would show backdoors for the government. That seems plausible on the face of it. Or am I being too much of a conspiracy theorist? Half your comment is just ad hominem. He's mentioned no 'sources' in this article, and it's not clear why we should be 'worried'. It's not a great article, but your response is overblown.
- pilom 15y agoFrom a government network security consultant: The vendors recommend disabling IPv6 because they understand that it is hard enough to harden just IPv4. By disabling IPv6 on government networks, you minimize the number of places a security auditor can miss something thus hopefully improving overall security. IPv6 really does require a new set of skills to secure so it is better to be old, simple, and proven (or at least we know where most of the problems are), than new, advanced and unknown when you are talking about security.