JWS- Verify failed with RSA public key but working with private key

[deepan-s]

I tried one of your example for JWS signing and verification.

I'm able to perform Sign with private key properly.
But I get below issue while Verify with RSA public key

Uncaught (in promise) TypeError: Cannot read property 'then' of undefined
at WebCryptographer.verify (jose-jwe-webcryptographer.js:316)
at jose-jws-verify.js:223
at Array.forEach ()
at JoseJWS.Verifier.verify (jose-jws-verify.js:183)

Code snippet -

var cryptographer = new Jose.WebCryptographer();
cryptographer.setContentSignAlgorithm("RS256");
var signer = new JoseJWS.Signer(cryptographer);
signer.addSigner(signer_rsa_key).then(function() {
        var message = document.getElementById("jws_text").value;
        var verifier = new JoseJWS.Verifier(cryptographer, message);
        verifier.addRecipient(rsa_key1).then(function () {
                verifier.verify().then(function (verified) {
                        document.getElementById("jws_text").value = 'Verified: ' + JSON.stringify(verified, null, 4);
                        }).catch(function (err) {
                            error.textContent = "verification failed: " + err;
                            }) });});

}

var rsa_key1 =
{
"n": "00:c2:4b:af:0f:2d:2b:ad:36:72:a7:91:0f:ee:30:a0:95:d5:3a:46:82:86:96:7e:42:c6:fe:8f:20:97:af:49:f6:48:a3:91:53:ac:2e:e6:ec:9a:9a:e0:0a:fb:1c:db:44:40:5b:8c:fc:d5:1c:cb:b6:9b:60:c0:a8:ac:06:f1:6b:29:5e:2f:7b:09:d9:93:32:da:3f:db:53:9c:2e:ea:3b:41:7f:6b:c9:7b:88:9f:2e:c5:dd:42:1e:7f:8f:04:f6:60:3c:fe:43:6d:32:10:ce:8d:99:cb:76:f7:10:97:05:af:28:1e:39:0f:78:35:50:7b:8e:28:22:a4:7d:11:51:22:d1:0e:ab:6b:6f:96:cb:cf:7d:eb:c6:aa:a2:6a:2e:97:2a:93:af:a5:89:e6:c8:bc:9f:fd:85:2b:0f:b4:c0:e4:ca:b5:a7:9a:01:05:81:93:6b:f5:8d:1c:f7:f3:77:0e:6e:53:34:92:0f:48:21:34:33:44:14:5e:4a:00:41:3a:7d:cb:38:82:c1:65:e0:79:ea:a1:05:84:b2:6e:40:19:77:1a:0e:38:4b:28:1f:34:b5:cb:ac:c5:2f:58:51:d7:ec:a8:08:0e:7c:c0:20:c1:5e:a1:4d:b1:30:17:63:0e:e7:58:8e:7f:6e:9f:a4:77:8b:1e:a2:d2:2e:1b:e9",
"e": 65537
};

rsa_key as per jose-jws-example.html but I have passed only n and e as I get these values only from RSA public key export.

Please guide me if I miss any to make it work.

[alokmenghrajani]

Are you using a browser or nodejs? Which version?

[deepan-s]

I'm using browser with latest version.

[alokmenghrajani]

Which browser? Version?

[deepan-s]

I tried it in chrome, firefox and edge as well.
chrome - Version 66.0.3359.139
firefox - 59.0.2 (64-bit)
edge - 38.14393.2068.0

[alokmenghrajani]

I think I can guess what's going on. When you sign the message, the kid is computed using n and d (probably a mistake? I'll have to figure that out). If you then verify with a key which doesn't have a d, you need to set the keyfinder argument (because the kid will mismatch).

We might want to change Jose.WebCryptographer.keyId to use n and e.

@rapropos, @rmhrisk, @csstaub: what do you think?

[deepan-s]

Thanks for your analysis.

Below the jsfiddle to replicate the issue. Maybe helpful for your further analysis.

https://jsfiddle.net/u0x2e2j8/5/
https://jsfiddle.net/grkz4akw/

[alokmenghrajani]

https://jsfiddle.net/grkz4akw/1/ shows you a workaround. We should also return a better failure message when the kid isn't found.

[deepan-s]

Thanks for the workaround. But the patch works only if we perform both sign and verify operation.

patch works only with sign operation https://jsfiddle.net/grkz4akw/3/

patch fails with verify operation https://jsfiddle.net/grkz4akw/2/

Any help, how we can apply this workaround for verify also and any plans on bug fix.

[alokmenghrajani]

https://jsfiddle.net/grkz4akw/2/ fails because of a trailing newline in message.

I was hoping to hear back from @rapropos or @rmhrisk, who wrote this part of the library. I'll give them a few more days and then roll out a fix.

[deepan-s]

Thanks. I will wait for the fix.

Just to know, I could not understand "trailing newline in message" which means I can fix this jsfiddle
https://jsfiddle.net/grkz4akw/2/ (I don't see any trailing newline in this example) or it is part of issue in library?

[alokmenghrajani]

See https://jsfiddle.net/grkz4akw/4/

[rmhrisk]

I think I agree with the conclusion, the problem is that this will be a breaking change (as has been pointed out above). I am not sure how large of an issue that will be for folks, maybe providing a helper to get the right kid based on the wrong kid if there was a good programmatic way to detect this wrong kid case?

[alokmenghrajani]

I plan to write a fix for this issue next week. Sorry it's taken this long, other things have been keeping me busy.