Add support for secp256k1

[OR13]

@dlongley @dmitrizagidulin I assume you may prefer for this to actually be added to crypto-ld. I'm eager to get this into DID Key, any quick advice that I might be able to follow to get a PR started on this.

We use https://github.com/bitauth/bitcoin-ts WASM module for browser support.

There are also:

• https://www.npmjs.com/package/elliptic

[dmitrizagidulin]

@OR13 - there's this repo, https://github.com/digitalbazaar/secp256k1-key-pair, which works with crypto-ld - would that help?

[dmitrizagidulin]

@OR13 so, unfortunately the fingerprint() function is not yet implemented - https://github.com/digitalbazaar/secp256k1-key-pair/blob/master/lib/index.js#L117 :) But at least the stub is there. (Same API as crypto-ld).

[OR13]

Seems like it uses: elliptic

Sounds like crypto-ld is ready to go (minus the fingerprint).... If I convert the key to jwk... I can calculate the fingerprint correctly... there is no standard for calculating a fingerprint on a base58 encoded key...

should I use the kid JOSE implementation for the fingerprint despite the fact that its using a base58?

Any chance I can use jwk formatted keys for secp256k1, use that as the fingerprint and just support publicKeyBase58 as an additional property?

Obviously, keyAgreement won't be included here.

[dmitrizagidulin]

I don't really have a strong opinion on this. @dlongley ?

[dlongley]

Note that a "fingerprint" in this case is the full public key or it won't work with did:key. If that's the fingerprint we're talking about here, I don't really care how it's calculated and prefer a standard or defacto standard that everyone is already using -- so long as it's fully compatible with the multifactor/multibase stuff, etc.

[OR13]

Thats not correct ATM i think...

https://github.com/digitalbazaar/crypto-ld/blob/master/lib/Ed25519KeyPair.js#L363

Note that ByHnpUCFb1vAfh9CFZ8ZkmUZguURW8nSw889hy6rD8L7 != z6MkqRYqQiSgvZQdnBytw86Qbs2ZWUkGv22od935YF4s8M7V

and similarly:

DmgBSHMqaZiYqwNMEJJuxWzsGGC8jUYADrfSdBrC6L8s !== zCDGPtoUB2xJzywzLodnst7ggAWY16JB9f7jxUUGebCaF5

{
  "@context": [
    "https://w3id.org/did/v0.11"
  ],
  "id": "did:key:z6MkqRYqQiSgvZQdnBytw86Qbs2ZWUkGv22od935YF4s8M7V",
  "publicKey": [
    {
      "id": "did:key:z6MkqRYqQiSgvZQdnBytw86Qbs2ZWUkGv22od935YF4s8M7V#z6MkqRYqQiSgvZQdnBytw86Qbs2ZWUkGv22od935YF4s8M7V",
      "type": "Ed25519VerificationKey2018",
      "controller": "did:key:z6MkqRYqQiSgvZQdnBytw86Qbs2ZWUkGv22od935YF4s8M7V",
      "publicKeyBase58": "ByHnpUCFb1vAfh9CFZ8ZkmUZguURW8nSw889hy6rD8L7"
    }
  ],
  "authentication": [
    "did:key:z6MkqRYqQiSgvZQdnBytw86Qbs2ZWUkGv22od935YF4s8M7V#z6MkqRYqQiSgvZQdnBytw86Qbs2ZWUkGv22od935YF4s8M7V"
  ],
  "assertionMethod": [
    "did:key:z6MkqRYqQiSgvZQdnBytw86Qbs2ZWUkGv22od935YF4s8M7V#z6MkqRYqQiSgvZQdnBytw86Qbs2ZWUkGv22od935YF4s8M7V"
  ],
  "capabilityDelegation": [
    "did:key:z6MkqRYqQiSgvZQdnBytw86Qbs2ZWUkGv22od935YF4s8M7V#z6MkqRYqQiSgvZQdnBytw86Qbs2ZWUkGv22od935YF4s8M7V"
  ],
  "capabilityInvocation": [
    "did:key:z6MkqRYqQiSgvZQdnBytw86Qbs2ZWUkGv22od935YF4s8M7V#z6MkqRYqQiSgvZQdnBytw86Qbs2ZWUkGv22od935YF4s8M7V"
  ],
  "keyAgreement": [
    {
      "id": "did:key:z6MkqRYqQiSgvZQdnBytw86Qbs2ZWUkGv22od935YF4s8M7V#zCDGPtoUB2xJzywzLodnst7ggAWY16JB9f7jxUUGebCaF5",
      "type": "X25519KeyAgreementKey2019",
      "controller": "did:key:z6MkqRYqQiSgvZQdnBytw86Qbs2ZWUkGv22od935YF4s8M7V",
      "publicKeyBase58": "DmgBSHMqaZiYqwNMEJJuxWzsGGC8jUYADrfSdBrC6L8s"
    }
  ]
}

This fingerprint implementation is custom (multibase / multicodec), its probably possible to implement it for secp256k1 and use it, we should just be really clear about how it works... it should be mentioned directly in the did key spec

[dlongley]

Sorry, to clarify, the fingerprint value doesn't need to match the base58 public key value byte for byte, it's just that the "fingerprint" (or value after did:key:) needs to be the multibase+multicodec encoded public key. If a hash of the public key were put there, the resolver could not produce a DID Document from it.

[OR13]

yes, i understand now after reading the source code. We should update the method spec to define the fingerprint behavior, and note that it matches: https://w3c-ccg.github.io/did-method-key/#format

[OR13]

Created an issue here: https://github.com/w3c-ccg/did-method-key/issues/10

[OR13]

Whats the significance of the 01 part of (multicodec ed25519-pub 0xed01 + key bytes) ?

https://github.com/digitalbazaar/crypto-ld/blob/master/lib/Ed25519KeyPair.js#L352

will this be the same for secp256k1?

[OR13]

Blocked by https://github.com/digitalbazaar/secp256k1-key-pair/pull/3

[dmitrizagidulin]

Whats the significance of the 01 part of (multicodec ed25519-pub 0xed01 + key bytes)?

It has to do with how multicodec encodes multi-byte varints, if I recall.

[dmitrizagidulin]

@OR13 I don't remember if it'll be the same for secp256k. Basically it's whatever the npm multicodec package will return for multicodec.addPrefix(), for the key type.

[OR13]

const multicodec = require('multicodec');
const buffer = Buffer.from('');
console.log(multicodec.addPrefix('ed25519-pub', buffer))
// <Buffer ed 01>

console.log(multicodec.addPrefix('secp256k1-pub', buffer))
// ???

No way to tell currently since the 'secp256k1-pub' change is not published yet....

[dmitrizagidulin]

@OR13 ah, boo, ok. I need to dig into https://github.com/multiformats/unsigned-varint carefully and figure it out, then.

[dmitrizagidulin]

@OR13 Ok, confirmed:

const varint = require('varint')
varint.encode(0xe7)
[ 231, 1 ]

which is the same as your implementation in https://github.com/digitalbazaar/secp256k1-key-pair/pull/3/files

buffer[0] = 0xe7
buffer[1] = 0x01

So, all good.

[OR13]

We still need to publish a version of https://github.com/digitalbazaar/secp256k1-key-pair before we can implement support in did key.

[dmitrizagidulin]

@OR13 secp256k1-key-pair npm package v1.1.0 is now published.

[OR13]

Blocked by: https://github.com/digitalbazaar/crypto-ld/pull/41

[OR13]

Blocked by: https://github.com/digitalbazaar/secp256k1-key-pair/pull/6

[dlongley]

To close this issue, we should add an example to the README. The new code base would not include anything directly.

[wip-abramson]

I would also like to use this library with secp256k1. But struggling to get it working using https://github.com/digitalbazaar/ecdsa-secp256k1-verification-key-2019

I am attempting this:

didKeyDriverSecp.use({
    multibaseMultikeyHeader: 'zQ3s',
    fromMultibase: createFromMultibase(Secp256k1KeyPair)
});

// This fails
const secpDidController = await didKeyDriverSecp.fromKeyPair({
    verificationKeyPair: secpKeyPair,
});

The above code fails. And every time I fix the current error, another one bubbles up.

The latest one I am stuck on is: Error: Cannot derive key agreement key from verification key type "EcdsaSecp256k1VerificationKey2019".

So basically wondering if anyone has an example of this working?