Cryptography May Not Be Dead, But It Is on Life Support

http://blog.varonis.com/cryptography-may-dead-life-support/

by rsobers • 13 years ago

57 45 13 years ago

blog.varonis.com

chris_mahan8• 13 years ago

Cryptography is very dead. Cryptography relies on the keys being kept private. The keys must be on computers to be usable. Computers can be compromised, therefore the keys can be compromised, therefore they encrypted payloads can be decrypted. Any question?

And by "Computers can be compromised", this means computers also cannot be trusted to hold any sensitive data.

valarauca1• 13 years ago

By this logic nothing is secret and nothing is safe, ever.

Horrible example. Crypto, like every other secret keeping method comes down to trust. Who, and what do you trust to keep your secret.

chris_mahan8• 13 years ago

You are correct: nothing is secret, nothing is safe. The governments can, at will, access any computer in the world, any file, and document. The only thing that stops them is the law, and as Snowden has shown, that's an oft ignored constraint.

So, one must assume that given enough interest, the governments will be able to compromise any system, and any data on these systems.

drdaeman• 13 years ago

One can make a microcontroller (not much space for government trojan horses here, huh) handle crypto. Or build a computer on FPGA.

It won't have much processing power, but it will still be sufficiently fast to handle strong crypto on text messages.

Or, if they're really really paranoid, there is always an option to use old trusty TTL components.

chris_mahan8• 13 years ago

And who is this microcontroller maker and how do I ensure that they have not been "cooperating" with anyone interested in breaking the encryption?

Or do I have to build my own chips? And hopefully I didn't make an implementation mistake.

TacticalCoder• 13 years ago

I don't totally disagree but... What about encrypting your message on a fully offline computer, like what Schneier is now doing if I'm not mistaken?

What if both him and his recipient do the same? Then you need physical access (and unnoticed physical access I'd add) to the offline computer to be able to decrypt right?

And physical access doesn't scale.

pyre3• 13 years ago

Devil's Advocate:

What if they swap Scheier's USB key (or drive) with one that will exploit the USB stack on his offline computer (disabling 'auto-mount' won't help you here)?

Yes, it's a physical access attack, but physical access to his person might be easier to achieve (while remaining unnoticed) than access to the offline computer.

chris_mahan8• 13 years ago

How would you decrypt the message? The only safe option would be to print out the encoded message, and retype it into the offline computer. I don't think anyone does that.

Also, the offline computer can be physically compromised.

EthanHeilman• 13 years ago

Its not cryptography which is dead, but the notion of a keeping a key secret forever. Key compromise will happen with some regularity, but there are three things you can do about it:

1. make key compromise harder, for instance require that an attacker compromise both your hardware and software to get your key,

2. limit the damage a key compromise can cause, for instance expire keys more often, randomly reissue keys,

3. and increase the speed with which you can recover from a key compromise. For example, the web PKI is completely compromised if a single CA key is compromised, but chrome can detect valid but forged google certs using key pinning and thereby alert the world. You can only use the stolen CA key to attack Google users once.

The security community is advancing on all three fronts.

seagreen• 13 years ago

You don't have to keep your private key on a general purpose computer. See Cardano -- http://trilema.com/2013/snsa-first-product-the-cardano/

chris_mahan8• 13 years ago

Cardano, you think, cannot be coerced by Romanian security services to backdoor their products? Also, check out the link at the bottom to polimedia.com, the parent company.

I don't think I would entrust my private keys to this outfit, sorry.

AnimalMuppet2• 13 years ago

This is not new at all. Clear back in "Applied Cryptography", Schneier said that there is no security without physical security. If someone can physically get at your box, it's all over. For most people, he said, that level of paranoia is overkill, but for some it is very real.

And then there's what happened to Target. Someone got to their machines, not physically, but with a trusted download. That paranoia now has to be carefully thought through by everyone who cares about their security...

codex• 13 years ago

Cryptography is not dead. Just because some motivated governments can theoretically attack your machine does not mean that cryptography is not useful, because for 99.9999999% of us, that outcome will never happen--and if it does happen, the information will not leak to those you want to keep it from (unless that's the government, of course). Cryptography still keeps us safe from all but three or four entities.

Put another way--the FBI can sneak into my apartment any time they want by picking my lock, but I still lock my door.

brazzy• 13 years ago

The article argues that your assumption that "advanced" techniques used by motivated governments now will stay the exclusive domain of those governments is wrong.

IT systems and networks have become so large and complex that vulnerabilities are everywhere. Your home computer may have a small enough attack surface by itself, but is constantly communicating with systems who don't - and much of your important data is kept in those anyway.

pyre3• 13 years ago

Some of those attacks are resource-intensive though. Only governments have the financing to carry them off, at least at the current time, and into the near-future.

ds9• 13 years ago

Cryptography is alive and well, but not because we can redefine the goals to disregard the most powerful adversaries. It's alive and well because there are ciphers that remain essentially unbreakable and are likely to remain so in the forseeable future.

The title is somewhere between misleading and dishonest. It's like saying that door locks are "dead" because burglars can go in the windows - in reality there's nothing wrong with the door locks, and the correct response is not to abandon the door locks but instead to secure the windows too (no pun intended).

The fact that there are side attacks is not a flaw of cryptography, rather it is a part of the environment in which it is practiced, and besides the attackers' workarounds are in principle subject to being defeated or avoided.

chris_mahan8• 13 years ago

One does not need to break the cyphers if one can obtain the decryption key, either by locating them, or "enhanced interrogation techniques" them out of biological systems.

Florin_Andrei• 13 years ago

for 99.9999999% of us, that outcome will never happen

It may actually happen for quite a bit of people. I think your percentage is about people where it makes a difference in their life whether or not the Three Letter Agencies decrypt their traffic or not.

FBT• 13 years ago

for 99.9999999% of us, that outcome will never happen

Note: As a fraction, the number you give for this is that it will happen to one in a billion people.

You are claiming that there are 8 people in the world that this outcome will happen to. While your general point may be true, I want to emphasise that this number is quite absurdly off. (It's off by at least two or three orders of magnitude.)

hydralist• 13 years ago

How about 10 factor authentication?

thrush• 13 years ago

Security is on Time Period Zero (Day Zero, and Year Zero didn't really seem to be the right terms). What we need in Crpyto is awareness and ubiquitous use. Additionally, it is not a fast moving concept. It will be long periods of sweat and churning before the internet is reasonably safe to use (I use the word reasonably, because anyone familiar with security understands that Perfect Security is not feasible. Rather we strive for security with negligible chance of failure).

If nothing else, Bitcoin as a protocol is a ray of hope for cryptography. If a digital crypto-currency ever becomes widely used, then crypto will become embedded in everyone's daily life (whether they're on Facebook and Gmail or not).

Public-Private Key Crypto is a concept that has only been in practice for less than 50 years. It has yet to be fully understood and implemented.

qwerta• 13 years ago

Cryptography is not dead, but useless. We are heading in direction where just possession of random data will be criminal offense.

eli• 13 years ago

How do you mean?

ReidZB2• 13 years ago

I'm not the OP, but:

Securely-encrypted data is supposed to be indistinguishable from truly-random data --- this is the widely-used definition of security for encryption. If laws are in place requiring that a user reveal any decryption keys upon a court order, then a truly-random file will "look" encrypted and so you may be ordered to provide your key to decrypt the file. Of course, it is truly random, and there is no key, so you would be in contempt of court. Indeed, assuming secure encryption, there is no way to prove that a truly random file is not actually encrypted.

In an extreme case, where the state outlaws secrets entirely, possession of truly-random data will look like you are attempting to hide secrets.

vezzy-fnord• 13 years ago

I absolutely detest FUD about quantum computing and how it's going to "destroy all cryptography; there's no use anymore", completely ignoring the fact that only asymmetric systems based on DLP and IFP will be compromised due to the efficiency of Shor's algorithm.

I always get particularly pissed because it makes the average person think it's no longer useful to encrypt their data, thus making them a vulnerable target. I wonder if it's a deliberately orchestrated disinformation campaign.

ReidZB2• 13 years ago

only asymmetric systems based on DLP and IFP will be compromised due to the efficiency of Shor's algorithm.

Quantum computing has larger implications than that. For instance, AES with 128-bit keys will need to be phased out since Grover's algorithm would allow a 2^64 time known-plaintext search. MD5 will likewise require 2^64 time to find a preimage, despite its resistance so far to non-quantum preimage attacks.

You should not discount that "only" asymmetric schemes based on DLP/IFP will be vulnerable, either. Much of our current key-exchange implementations do not have post-quantum strength at present.

Still, as you say, it is not "end of the world" status for cryptography. It will just require a lot of effort on the part of system builders to implement post-quantum constructions. The issue of backwards compatibility and widespread implementations will (as it always is) be a significant hurdle.

It's my experience, however, that the average person doesn't know or care about encrypting their data at all. It's almost certainly the opposite on a site like reddit or HN, but anecdotally, when my parents' friends (as an example) ask what I study and I say "cryptography," it's exceptionally hit-and-miss if they'll even know what I'm talking about. (A nontrivial portion of people think I've said "photography," actually!)

I've found that "encryption" is more widely-known than cryptography, but even then, it's been my experience that most (non-techie) people have no clue about it. But, I suspect that changes quite drastically given your social circle, area of the world you live in, and culture.

yeukhon2• 13 years ago

Disclaimer: I went to the conference (well it's actually my school). The followings are supposed to be addressed during the QA but I could only ask one question... sorry for the long baffling.

This is the golden age for cryptography, thanks to education and hardworking people.

People are actively attacking our cryptographic knowledge and our implementations. As controversial as it may sound, if it weren't all the active attacks on our cryptographic infrastructure, we probably will be okay with RC4 and MD5. Of course we know they are weak and they are not reliable.

So let's thanks everyone, including the state-sponsored attackers.

This is golden age because we long know that relying on mathematical hardness assumptions is not safe. Maybe a decade later someone discover a theorem to factor large number very efficiently and then boom all the encrypted communications using RSA will be broken. We are slowly moving from that kind of dependencies. We think there are better ways to solve our encryption. Much like in 20th century the arm race gave rise to active advancement in all disciplines of engineering and science, cryptography is also growing.

Thanks to all the attackers out there we now know it is important to teach everyone about computer and web literacy. We know this should be part of education. In addition, we must make tools more accessible to users. At #realworldcrypto 2014 someone said PGP has been around what two decades? Why hasn't everyone in the tech community using it? Why are my non-geek friends not using it? Why am I not using it? Servers that retain user data or transfer user data should all be over HTTPS now. Implement 301 redirect on http end points and on HTTPS endpint add HSTS header. Implement Content-Security-Policy to harden what resources can be loaded on your website. Add X-Frame-Options to control whether you want your site to be frame/iframed or not. I can go on and on but you get the point. This is a long battle and not easy to fix.

Cryptography is not dead. What is dead is our assumption that we can rely on assumptions and that kind of dependency is going to harm us some time in the future. For how many more years? We don't know. It is possible no one can ever come up with an efficient algorithm to break factoring.

Yes. One problem in cryptography has to do with the key storage. I see that in the future HSM will be cheaper and people can enjoy that as opposed to a plaintext file in your $USER/.ssh/ directory. Look, cryptography is not silver bullet. You can't eliminate people from making mistake, but we can look at what things can be improved to make mistake fail quick and safely. Idea? Maybe instead of one key, we have multiple partial keys stored on multiple servers? But key management and key synchronization is going to be a headache. And look, if someone inject a malware in the network and has some insight knowledge of the network, there is very little you can do.

Never confuse NSA revelation means we must implement things so secure that we can't even tell Bob is Bob. We can't have 100% anonymity and we can't enforce that. The world needs interaction. The ability to choose is the right direction and I hope companies will start to realize that we don't live in the 80s anymore.

The hardest problem to solve is to tell whether the server is doing what it is said. People are working on verifiable search but what about whether site is actually hashing your password? Client-side encryption is important and mufti-identity remain to be solved. Personally, I'd like to see Persona widely used so I can just set up my own federated authentication server to authenticate my own email.

Again, as controversial as it may sound, knowledge exists because we can think and because we can think we have desire and goals. Knowledge doesn't grow out of the trees. The are always accidental and incidental. We don't start inventing things out of the thin air. I like the idea of knowledge as Yin-Yang. We don't start having cryptography because there is such a thing called cryptography. Because we want secret to be hidden and safe from evedropper, we invented substitution cipher schemes. Because we now have digital communication and we need to prevent MitM we need a better cryptography and this is why RSA and DHE are useful. We know SHA is never meant for hashing password because it's fast so we invent other kinds of cryptographically hard hashing algorithms like bcrypt and scrypt. If it weren't Miller's paper on fuzzing, we probably would neglect fuzzing testing and our unix command line tools will probably continue to fail hard. If it weren't NSA, how many of us would ever pay attention to the problem in OpenSSL and RNG? There is always a constant Yin-Yang interaction in the pursuit of knowledge. One nice property of security proof is that we always have to model the evil in our proof construction...

saraid2162• 13 years ago

Never confuse NSA revelation means we must implement things so secure that we can't even tell Bob is Bob. We can't have 100% anonymity and we can't enforce that. The world needs interaction. The ability to choose is the right direction and I hope companies will start to realize that we don't live in the 80s anymore.

I want to quote this for emphasis.

One of the huge mistakes that drives our current despair is the unconscious ideal of a world where we aren't significantly interacting except through anonymous transactions. This isn't tenable. More importantly, this isn't human. Privacy is less about having security and more about convincing other people to avert their eyes. That's why the fault is on the NSA, not on, say, messaging protocols. Privacy is not sneaking into someone else's home. The burden of maintaining privacy is on the person who could violate it.

StandardFuture• 13 years ago

The burden of maintaining privacy is on the person who could violate it.

I want to quote this for emphasis. ;)

twobits• 13 years ago

"We can't have 100% anonymity and we can't enforce that."

We can and should have pseudonymity.

"That's why the fault is on the NSA [..] The burden of maintaining privacy is on the person who could violate it."

Then we should abolish the police too. Criminals will just feel bad knowing they are at fault. Who is to blame, is irrelevant to how you address a problem.

nullc• 13 years ago

Nothing is absolute, you might be hit by a meteor in the next moment. So what that there is a non-zero chance that if you're targeted by the right parties you may not be able to keep secrets?

Moreover, cryptography is more than just secret keys being kept secret.

For an example of cryptography which doesn't depend on secrecy, consider SNARKs (succinct non-interactive arguments of knoweldge): E.g. I can run a program and give you its output along with a compact proof that the output was the faithful output of the program. The size and complexity of verifying the proof is only a product of the cryptographic security level. Given cryptographic assumptions it is computationally infeasible for me to generate a fake proof.

The ability to prove the validity of execution in basically no more time than it takes the read the program being verified is a very powerful result of cryptography which doesn't depend on secrecy.

(The most efficient constructions of this currently need some secret data, but it's not a fundamental requirement)

greenyoda• 13 years ago

"They had a system that looked for phones moving towards each other, turned off and then turned on when they are turning away from each other. Looking for secret meetings."

And that probably wouldn't require a lot of computation either, since very few people would turn their phones off while going somewhere and then turn them on again. So the people who are doing this are smart enough to know that they can be tracked by their cell phones, but not smart enough to know that turning them off and on again calls a lot of attention to themselves.

So if you're going to a secret meeting that you don't want the NSA to know about, leave your phone at home.

agreenjay2• 13 years ago

A few of my thoughts. Bruce did make an important point during his presentation, which I didn't write about in the piece. Whenever possible the NSA goes around crypto and other security. The agency has immense power--through, for example, QUANTUM Insert and redirecting to Foxacid servers--to target individuals. The real-issue is BULK collection, and that's where strong crypto will help when placed on endpoint devices and of course at the backend. In this sense: if we make it more expensive and more difficult to get at the data, the NSA will go after easier targets--this is kind of implied in some of the Snowden documents according to Bruce. So crypto is "less important" because if they really, really want to governments and next-gen cyber thieves will get at some of the data. But there are a lot of conventional techniques--better authorization controls, and constant monitoring of activity-- along with encryption that can limit the amount of data that's exposed.

api• 13 years ago

"Cryptography is dead" is a weird hypey framing. The problem isn't with crypto but with the underlying system... crypto is only as secure as the place it's performed.

ChuckMcM• 13 years ago

Cryptography is not on life support any more than Schlage door locks are "dead" because they can be picked by anyone with the tools and training to do so, and kicked in by anyone able to bring more than 50 lbs to bear on the lock.

When I've given security talks I've always tried to compare things like air travel, which isn't 100% safe (planes do fall out of the sky) but is 100% "worth it" for most of us. And most of us won't be the focus of a government investigation.

So the reasoning that we can't make systems 100% secure, therefore security is dead, doesn't persuade me.

What will be interesting though is the use of cryptography to protect things like our pin numbers at the checkout line. That is an area that needs improvement right away.

yeukhon2• 13 years ago

What will be interesting though is the use of cryptography to protect things like our pin numbers at the checkout line.

Once my professor said the ATM machine in Japan randomized the order of the keys on the screen. I don't know how much effort it will require someone to upgrade the software to have a randomized keypad on the screen, but the risk of upgrading it is worth it.

agreenjay2• 13 years ago

Thanks for this comment. This is probably closer in spirit to Schneier's view. And his big point is that we can do more to prevent mass collection of data through strong cyrpto-- a good door look and bolt. It appears to me -- I was at the conference and wrote the article--that many in the community are surprised/impressed by how effective APTs and other techniques--packet injection-- are at getting around the lock.

Cryptography May Not Be Dead, But It Is on Life Support

Related Stories