Use the correct verifier for RSA PSS scheme keys

[rdimitrov]

The following PR fixes an issue with go-tuf's support for RSA key types.

Details:

Apparently for RSA key types signature.LoadVerifier() is defaulting to returning a verifier that is RSAPKCS1v15Verifier which corresponds to the RSA PKCS#1 scheme.

Based on the latest TUF spec, TUF uses RSA PSS instead of PKCS#1. This manifests itself by causing go-tuf v2 TUF clients to fail metadata verification in case of RSA key types.

The fix for this is to invoke the proper verifier for the PSS scheme. This means that when we are verifying a metadata signed by an RSA key type, we should use LoadRSAPSSVerifier() instead of the generic LoadVerifier().

References:

• https://theupdateframework.github.io/specification/latest/#file-formats-keys
• https://github.com/sigstore/sigstore/blob/a153736bf87443389feee6aa787e97b81839175d/pkg/signature/verifier.go#L41

Credits:

Thanks to @kairoaraujo for reaching out and helping track this together! 🚀 Cheers! 🍻

What's left:

• Update the tests

[kairoaraujo]

I've tested it, and it solves the issue 😃

[jku]

Based on the latest TUF spec, TUF uses RSA PSS instead of PKCS#1

Note that these are just examples essentially, definitely not an exhaustive list of the only schemes that can be supported: I would avoid removing support for the pkcs1v15 scheme if possible as removing optional features can be a real head ache for users...

EDIT: the PR does not seem to remove the support, I just looked at comments originally and misunderstood

[jku]

Just as an idea... I would love a keytype support test in tuf-conformance: theupdateframework/tuf-conformance#159 -- implementing that would give you keytype test coverage "for free" in go-tuf and should not be more than 20 lines of code.

EDIT: I did an initial test in tuf-conformance theupdateframework/tuf-conformance#167

[kommendorkapten]

FYI: I'm working on this, should have a working update within a few days.

[rdimitrov]

@kommendorkapten - Awesome! 🙏 I'd consider this my Christmas gift 😄

[kommendorkapten]

A lot of the tests are working, I'll hunt down the rest failures tomorrow I hope

[kommendorkapten]

This is now ready for review, the tests were failing as they were signed using PKCS#1 v1.5 but the key scheme said rsa-pss-sha256, I resigned the test data (with the helper tool added to this pr), and then made sure any signing and verification in the tests uses PSS. And when comparing to/from bytes I ignore whitespaces as its brittle to care about them.

[rdimitrov]

🚀

[MDr164]

Works locally as well for me. I'd say it's fine to go in. There are a some smaller nits in there but none of them affect the functionality so I'll refrain from blocking a merge :)

[rdimitrov]

Thank you all for reviewing and contributing to this! 🙌 💯