Add RS512, PS256, and PS512 support

Zaubrik / djwt

Create and verify JSON Web Tokens (JWT) with Deno or the browser.

MIT License

228 stars 23 forks source link

Add RS512, PS256, and PS512 support #41

Closed invisal closed 3 years ago

invisal commented 3 years ago

With new release of god_crypto v1.4.7, you can now add support for the following algorithms: RS512, PS256, and PS512.

RS512

await rsa.sign(message, { algorithm: "rsassa-pkcs1-v1_5", hash: "sha512" })
await rsa.verify(signature, message,  { algorithm: "rsassa-pkcs1-v1_5", hash: "sha512" });

PS256

await rsa.sign(message, { algorithm: "rsassa-pss", hash: "sha256" });
await rsa.verify(signature, message, { algorithm: "rsassa-pss", hash: "sha256" });

PS512

await rsa.sign(message, { algorithm: "rsassa-pss", hash: "sha512" });
await rsa.verify(signature, message, { algorithm: "rsassa-pss", hash: "sha512" });

timonson commented 3 years ago

Hi @invisal thank you for this update!

Unfortunately, the tests for the new algorithms fail. If you don't mind please take a look at this new branch https://github.com/timonson/djwt/tree/alg-patch1 and run deno test -A.

I got the JWTs for comparison from https://jwt.io/.

An additional test for PS512 is missing but the tests for RS512 and PS256 fail because the signatures don't match.

invisal commented 3 years ago

Thank. I have checked the problems and I found the problem in my implementation:

For RS512, the encryption is right, except that I write the wrong oid (using sha256 oid). I use verify code to test the sign code and my verify code does not include checking oid. I will update more test case for this one.
For PS256, there is small error in my code as well and add more test cases.

The test method for testing PS256 in your code is also wrong. PS256 and PS512 include random salt. That's mean that it always produce different signature for same message.

I will patch the fix this week and I will update back to you on this issue.

itohatweb commented 3 years ago

Maybe add support for ed25519 too?

invisal commented 3 years ago

I have updated the fix in 1.4.8. However, it still fail the PS256 test case because PS256 is random algorithm. It always reproduce different signature every time you sign same message.

@itohatweb elliptic curve algorithm is also in my roadmap. But I need to do more reading to implement it.

timonson commented 3 years ago

The tests pass now.

I would love to get a code review before I merge these changes: https://github.com/timonson/djwt/pull/44