ring
ring copied to clipboard
briansmith
NIST curve P-521 SHA-512 support for generating keys and signing
I'd like to have support for ECDSA_P521_SHA512_ASN1_SIGNING signing, DER export and key generation as there already is support for ECDSA_P256_SHA256_ASN1_SIGNING or ECDSA_P384_SHA384_ASN1_SIGNING.
See also: https://tools.ietf.org/html/rfc5480#section-4 for a list of possible combinations.
Which software uses p521 SHA-512?
I just wanted it for completion. If you don't feel there is a need, feel free to close this.
Openssl supports p521 but it seems to sign using SHA-256. E.g. openssl genpkey -algorithm EC -pkeyopt ec_paramgen_curve:P-521 -pkeyopt ec_param_enc:named_curve generates a key but if you use it to sign a csr, it uses a SHA-256 hash. Firefox is at least capable of displaying the key, haven't tested more.
I would also like this feature, I'm trying to implement a software which is built on p521 SHA-512 and would like to use ring as a crypto library. This is a feature that cfssl does have.
I would be interested in donating to help get this feature implemented. I went ahead and sent you an email @briansmith.
ecdsa_secp521r1_sha512 is also one of the supported signature algorithms in TLS 1.3: https://datatracker.ietf.org/doc/html/rfc8446#section-4.2.3
P-521 is used in JOSE, COSE, as well as supported in Web Cryptography API. Developers may choose to use this curve in a variety of applications for quite some time now.
It also gets used in Terraform plugin handshake validation which admittedly, are mostly in Go, but I ran into issues recently building one in Rust because of this lack.
I think P-521 is also important from the post-quantum perspective. Are there any thoughts, plans for the implementation?
I think P-521 is also important from the post-quantum perspective.
Note that P-521 (and all the other standard ECDSA curves) are not quantum resistant. Quantum-resistance will require a completely different signing algorithm (not ECDSA or EdDSA). See https://csrc.nist.gov/projects/post-quantum-cryptography for more information.
FWIW, When I asked "Which software uses p521 SHA-512?," what I meant was "What software requires you to implement P-521, given that P-256 and P-384 are already available to you?" Of course I know of all the standards that provide ways to use P-521, but I don't know of any popular software that has a P-521-only configuration.
Regardless, I am more open to supporting P-521 (and P-521 SHA-512 signature algorithms) than I was in the past.
Just, I'm surprised so many people care about it. I would expect to find a lot more X448-only software than P-521-only-software.
Of course I know of all the standards that provide ways to use P-521, but I don't know of any popular software that has a P-521-only configuration.
As near as we can tell, this software only has support for P-521. https://cert-manager.io/docs/projects/csi-driver-spiffe/
I have another use case as well. Our organisation uses Cloudflare Teams and it's root cert (secp521r1 with SHA-512) cannot be verified by rust-webpki as it relies on ring.
@nazure could you attach a copy of the root cert here? I'd be interested.
Yup, sure. Here's the official link to the certificate: https://developers.cloudflare.com/cloudflare-one/static/documentation/connections/Cloudflare_CA.pem
Thanks for looking at this.
@nazure interesting, indeed it seems to use P-521 cryptography with SHA-512. For full transparency, I probably won't contribute code to ring, but I asked to figure out the way that the usage occurs.
For anyone curious, it looks like Vlad did nearly complete this - https://github.com/briansmith/ring/pull/1631
To clarify the above: Vlad's work in PR #1631 will add P-521 signature verification support, whereas this issue is for the (private) keypair generation, keypair parsing, and signature generation. That is, Vlad is working on the public key operations, whereas this issue is about the private key operations, i.e. ones that would be needed by rcgen. I filed issue issue #1771 to track the work that's blocking the signature verification work.
In terms of supporting the private key operations tracked by this issue, I think we're going to block this on #1771 being completed.