There Is No Diffie-Hellman but Elliptic Curve Diffie-Hellman
keymaterial.net
keymaterial.net
It then turns out that if you search for the algebraic groups that you can use for DH, you end up looking for candidates within the 2 possible kinds of algebraic groups: linear algebraic groups and abelian varieties. Within the abelian varieties, it turns out that elliptic curves are both the simplest kind, while also being perfect for DH. Within the linear algebraic groups, the possible cyclic groups are just the additive and multiplicative ones: The additive ones are hopelessly insecure, while the multiplicative ones are just the unit groups of a field, which are special cases of elliptic curves anyway. So elliptic curves really are all you need for DH.
Like, using something like SageMath for algebraic structures, a prover like Lean to verify properties — to get a feel for how things actually fit together.
There's something cool about trying to reimplement known standards just to understand them better (with the usual "don’t roll your own crypto" warning, of course). By the way, it would be nice to have a kind of sandbox for the more formal path to explore how to build my own SP-network, prove that the Decisional Diffie-Hellman assumption holds or something.
I've seen cryptopals and cryptohack.
eg CryptHOL needs a ton of infrastructure, and in still very limited: https://eprint.iacr.org/2017/753.pdf
If we do not talk about modern exotic stuff (post-quantum crypto, zkSNARKS, homomorphic encryption), the 99% of everyday cryptography is based on two building blocks:
1. Symmetric crypto for ciphers and hash functions.
2. Algebraic group with "hard discrete log problem" for key exchange, signatures, asymmetric encryption and simple zero-knowledge proofs.
Historically, these two categories are filled with a zoo of protocols. E.g. AES is a block cipher, but SHA(1,2) is a hash function.
Today, you can roughly achieve everything of the above with two universal building blocks:
- Keccak for all of symmetric crypto: it is suited both for encryption, hashing, duplex transcripts for ZK protocols etc.
- Ristretto255 group based on Curve 25519: for diffie-hellman, signatures, key derivation, threshold schemes, encryption and more.
The problem is that none of the described features is implemented in a turnkey standard, and we are still stuck using older crypto. Heck, even Git is using SHA-1 still.
Then, after you have your building blocks, there are more hairy stuff such as application-specific protocols: TLS, Signal, PAKE/OPAQUE, proprietary hardware security schemes for full disk encryption and access controls etc.
Usually it goes like that: someone made something useful optimised for a specific use-case with certain time (or competence) constraints, within a total lack of decent alternatives. Then people adopt and use it, it becomes the standard. Then people want to do more things with it, and try to build around that thing, or on top of that thing and Frankenstein monsters get born and also become standard.
If you start from scratch you can do a crypto protocol that is both better designed (causes less UX pain and critical bugs) AND performs better on relevant hardware. Also do not forget that performance is easily solved by hardware: Moore's law and then custom hardware extensions are a thing.
Example: Keccak is so much better from the composition perspective, that when used ubiquitously you'd definitely have ubiquitous hardware support. But if everyone continues to use a mishmash of AES and SHA constructions on the pretext of "Keccak" is not as fast, then we'd never move forward. People would continue building over-complicated protocols, bearing subpar performance and keeping the reputation of dark wizardry inaccessible for mere mortals.
"Just write slow algorithms, hardware will eventually get faster" doesn't really work when talking about performance implications now. If the hash algorithm used million of times doesn't perform on current user hardware, then the algorithm is simply not a good fit.
> and then custom hardware extensions are a thing.
That's the kind of trade-off I eluded to as well. As a developer of a tool (e.g. git), I'd pick hash algorithms that do have hardware extensions on the most common hardware and not use something, that may eventually get hardware extensions.
I guess developing such protocols right now for the future might still be advisable, but it seems odd to critic software that use well-optimized algorithms and fulfill the requirements.
Example: Git users do need both corruption protection AND secure authentication. If authentication is not built in, it will have to be built around. Building around is always going to be more costly in the end.
Unfortunately, 20-30 years ago considerations such as "sha1 is shorter + faster" were taken seriously, plus all the crypto that existed back then sucked big time. Remember Snowden scandal in 2013? That, plus Bitcoin and blockchains moving towards mainstream brought about review of TLS, started SHA-3 competition. Many more brains turned to crypto since then and the new era began.
The problem is... it doesn't line up with the facts.
Git has been using SHA-1 hashes for signatures since very early on. It also has claims in its documentation about "cryptographic security". It does not rigorously define what "cryptographic security" means, but plausibly, it should mean using a secure hash function without known weaknesses.
"So that was one of the issues. But one of the issues really was, I knew I needed it to be distributed, but it needed to be really, really stable. And people kind of think that using the SHA-1 hashes was a huge mistake. But to me, SHA-1 hashes were never about the security. It was about finding corruption.
Because we’d actually had some of that during the BitKeeper things, where BitKeeper used CRCs and MD5s, right, but didn’t use it for everything. So one of the early designs for me was absolutely everything was protected by a really good hash."
https://github.blog/open-source/git/git-turns-20-a-qa-with-l...
Also, SHA-1's preimage resistance (which still isn't broken) is necessary for the security of signed commits, regardless of the hash function used for the signature itself, since a commit object references its tree and predecessor commit by their SHA-1 hashes.
Unfortunately, Keccak and sponge constructions in general are inherently sequential. Even with hardware acceleration it heavily restricts possible performance. For example, AES-CBC encryption is 4-8 times slower than AES-CTR on high-end CPUs with AES-NI available. VAES makes the difference even bigger. Algorithms like AES-GCM, ChaCha20, and BLAKE3 are designed specifically to allow parallelization.
CShake128 is much better replacement for hmac and sha512 in (zk)proofs, while Kangaroo for things like FDE and massive volumes of data.
Couldn't you simply using BLAKE3 instead? To my knowledge BLAKE3 exactly was designed to solve this "parallelism problem" by combining the "cryptographic ideas" of BLAKE2 with the binary tree structure of Bao (the latter was designed to make the hash construction easy to parallelize).
https://github.com/nakov/practical-cryptography-for-develope...
OTOH if you want to win a Turing Award…
That may surprise people but we have no proof of the robustness of the encryption we use. In fact it might not work! Consider a one time pad for really extreme security needs. One time pads do have a proof (see the work of Shannon).
The symmetric encryption used for the vast majority of actual encryption has always been safe. Grover's algorithm does not scale appropriately (exponential on key length) unlike Shor's and cannot be parallelized. Worst case, AES-128 becomes as week as AES-64. So we bump it to AES-256 and we're back at the same difficulty as before (impossible). There's a degree of confidence this will always hold true.
Like assuming f(x) is hard, let us try to prove these other properties of security.
E.g. "Assuming AES block cipher is a perfect pseudorandom function, prove AES-GCM construction is secure up to certain number of messages"
You can find more details in Silverman's book on elliptic curves, or if you don't have access to that, see Section 24.7 in Lecture 24 of https://ocw.mit.edu/courses/18-783-elliptic-curves-spring-20....
One could nitpick by pointing out that singular cubics are not elliptic curves (which are, by definition, smooth projective curves). One could also nitpick that the real reason you don't want to use DH in the multiplicative group of a finite field is that there is a known subexponential time algorithm that breaks DH in that setting (forcing > 3000-bit key sizes) whereas for elliptic curves the only known attacks take exponential time (and 256-bit keys are fine). But I don't think either of these nitpicks contradicts the author's thesis.
Another point that the author glosses over a bit is that higher dimensional abelian varieties offer other instances of DH that are genuinely different from ECDH, and that are occasionally useful (mostly the case of Jacobians of hyperelliptic curves of genus 2). There isn't really a trick to make an arbitrary hyperelliptic curve DH/abelian variety DH instance a special case of ECDH: if anything, the relationship would be in the reverse direction.
ECC also used to have patents which restricted adoption back in the day.
There also used to be a lot of conspiracy theories that NSA backdoored nist curved ( e.g. https://m.slashdot.org/story/191445 ). Probably fud, but it slowed down adoption of eliptic curves.
Irrespective of the curve issue, ed25519/x25519 is superior and has other nice properties like not catastrophically breaking if you can't generate a unique random "k" for ECDSA as PlayStation discovered the hard way[1].
How does that work? I thought that this vulnerability was independent of the curve that is used. Ed25519 has a bunch of other nice properties related to resistance to programming errors, but I never heard of this one.
The trick is that this value in Ed25519 is a deterministic hash of the private key and the message, which means that for the same message, the same value will always be chosen, and since the private key is part of the hash, for different messages the value is basically random unless you already know the private key.
Yes, the same trick can be used with any curve, and it will avoid the vulnerability; but for Ed25519, it's always been part of the specification so everyone does it that way, while for ECDSA it's a later addition (RFC 6979) and many implementations still do it the old way.
Not FUD at all. NIST revoked their recommendation to use those curves because of it. The facts are that a "backdoor" exists whether anyone knows what it is or not, and NSA could not explain how those particular numbers were chosen so out of caution we must assume they know the backdoor and not use those curves.
The P-series curves are still a NIST recommendation and widely deployed in TLS. The B-series fell out of fashion due to practical reasons.
But for some reason the author keeps referring to the underlying group as "Diffie-Hellman" instead of the process operating on some group.
The result is interesting, the journey to get there was very confusing.
I know KaTeX supposedly lets you copy-and-paste the text, but that won't get you text that is fit for any purpose.
And if not, then that seems to be something that is now within reach of modern AI and just needs application.)
KaTeX will also style the text so that it visually resembles the image, stopping this problem from occurring for non-blind readers. You could have a screen reader that parsed all of the layout, but it seems like you might just as well OCR the image at that point.
Rather:
- The background of the math images is white, which doesn't match the background of the website;
- The baseline of the text in the images is not aligned with the baseline of the surrounding text.
I was imagining a more-competently-put-together website.
For any n, there is only one finite cylic group of size n, up to isomorphism (namely Z/nZ). As the author mentions, there are concrete choices of finite cyclic groups (like Z/nZ) that make Diffie-Hellman insecure. Consequently, the security comes entirely from the choice of the _representation_ of the group elements (since the algebraic structure is the same, i.e., it is just a relabeled Z/nZ).
In a similar spirit, people sometimes argue that Diffie-Hellman in some group G must be equally secure as in another group H, because the groups are isomorphic. This is unsound. In order to make such an argument sound, one needs to prove (or at least mention in case it is trivial) that one can _efficiently_ compute the isomorphism and its inverse.
But I’d agree that it’s not standard notation to use 0 for a terminal object in an arbitrary category. I’d guess that most people use 1 instead, so that for example 1 × X ≅ X. (The post talks about group objects in the category of algebraic varieties (over some field), in which case 1 seems to be more appropriate than 0.)
To me this sounds a bit like an a-posteriori justification of why people use elliptic curves to do cryptography. This feels weird to me, as elliptic curves are a well-studied subject in algebraic geometry, whose history reaches back over 150 years to Clebsch. So when Victor Miller and Neal Koblitz proposed to do cryptography with them, they already knew very well about elliptic curves (both have a background in algebraic geometry of finite fields).
There was a period during which there was lots of interest in using abelian varieties of higher dimension (arising as Jacobians of curves of higher genus), with dimensions g=3 and g=4 being particularly attractive because then you could work over a very computationally friendly base field like Fp with p = 2^61-1. But it turns out the discrete logarithm problem (and therefore DH) is strictly easier in these settings (one can exploit Weil restrictions to get an algorithm that is still exponential-time but strictly better than O(p^(g/2)). But this wasn't known until the 2000's.
That leaves g=1 and g=2 as the best choices, and the group law is faster and simpler for g=1, and as far as I know nobody is really working on the g=2 case anymore (but there was a lot of activity in this area 10-20 years ago).