Inconsistent signature length on ECDSA

[dav1app]

Hey,

I am building an hybrid application that shares the same code on both server and client sides. Same JS files, same modules. After some hours trying to figure out why I couldn't use WebCrypto -> Export ECDSA -> Send to the server -> Check the signature, I discovered that the signature has an inconsistent byte length and aways start with the very same bytes. This is the buf2hex read of it:

I found this pretty weird and almost sure that this should not happen. I run the same code on the browser to check if this behavior is normal. Here the results:

As I said before, this is the same code running. You can notice the difference. The code is the following. The "octano" module is just the layer to make the code runs on the browser and server. It keeps the same parameters as the original functions.

function test(){
  var publicKeyExported = {
    "crv": "P-256",
    "ext": true,
    "key_ops": [
      "verify"
    ],
    "kty": "EC",
    "x": "oVlNnOyWWwcIfhd73uMLrldbAy2YMRtuTjIY1Xz-I1o",
    "y": "RLMChUc4EOuWcjJcFr2knwxVsIiaTtLPsNRMFLK0ku4"
  }

  var privateKeyExported = {
    "crv": "P-256",
    "d": "VNLMIqrU9iBgPZIkxVrtIsB4wL6-lRE0e0SRNm0LeVY",
    "ext": true,
    "key_ops": [
      "sign"
    ],
    "kty": "EC",
    "x": "oVlNnOyWWwcIfhd73uMLrldbAy2YMRtuTjIY1Xz-I1o",
    "y": "RLMChUc4EOuWcjJcFr2knwxVsIiaTtLPsNRMFLK0ku4"
  }

  var algoDefine = {
    "name": "ECDSA",
    "namedCurve": "P-256"
  }

  var signAlgo = {name: "ECDSA", hash: {name: "SHA-256"}}

  var dataToSign = 'test'

  return Promise.all([
    octano.util.importKey("jwk", privateKeyExported, algoDefine , true , ['sign']),
    octano.util.importKey("jwk", publicKeyExported, algoDefine , true , ['verify'])
  ])
  .then(x => {
    return octano.util.signData(signAlgo, x[0], octano.util.textEncoder(dataToSign))
    .then(x=>{return octano.util.buf2hex(x)})
  })
  .catch(e => {console.log(e)})
}

test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})
test().then(x => { console.log(x)}).catch(e => {console.log(e)})

INB4: Not the font, I use monospace. IINB4: Not a problem on the buf2hex, I've tested it in another projects.

[thelunararmy]

This appears to be the same problem as outlined in #22 , we are looking into it.

[thelunararmy]

Main problem we identified was node's crypto library uses openssl to sign data, which results DER encoded openssl signatures which intrinsically contains metadata. What we will need to do is deconstruct the DER signature, extract the signature value, pad it properly, and return it.

[amark]

@thelunararmy @EternalDeiwos @davimello28 also related to https://github.com/anvilresearch/webcrypto/issues/74 that could probably be merged into this thread (or #22 ).

Any timeline on when this can be fixed? (In all honesty, of course, I know we're all busy)

[amark]

OSSL might have a nifty function for this already, here ConvertWebCryptoSignatureToDerSignature at https://github.com/PeculiarVentures/node-webcrypto-ossl/blob/master/src/ec/ec_dsa.cpp#L175 ? Is this theoretically what is needed? Or not related?

[rmhrisk]

As a FYI, we have created a purse TS Webcrypto polyfill also - https://github.com/PeculiarVentures/webcrypto this removes the C++ build-time dependencies. It is not full featured enough to replace our current node-webcrypto-ossl for all use cases due to dependency limitations with this approach but it may be useful for those who are encountering the problem discussed here in this bug.