Bug report: PGP Sign operation creates encrypted output, while it should be signed cleartext.

[engrach]

The PGP Sign function currently uses the box function from kbpgp. The box-function is meant for encryption. While it is functional it is not creating the correct output. The clearsign function from kbpgp should be used instead. This will create an RFC4880-compliant message with human-readable text and a PGP signature.

To Reproduce
Select the PGP Sign funtion.
Supply a private key and passphrase
Add some random input

The result is a PHP message with the header:
"-----BEGIN PGP MESSAGE-----"

Using the clearsign will create a message with the header:
"-----BEGIN PGP SIGNED MESSAGE-----"

Additional context
I have created PR #2696 to change the function from kbpgp.box to kbpgp.clearsign,

[engrach]

@GCHQDeveloper581
Currently the PGPSign uses the kbpgp.box function which is used for signing a file. Is that the purpose of this function?

There is a poorly documented kbpgp.cleartext function which can be used to created a human-readable signed message. This is the format used for RFC9116 security.txt. See paragraph 2.3 of the RFC.

Should this be replaced or should a PGP Cleartext sign operation be created?

[GCHQDeveloper581]

I was thinking about this today when I was having a look at your PR.

I need to think this through a bit further. The PGP operations are very tightly coupled with the kbpgp library and this differs slightly in terminology to the OpenPGP standard (as implemented by gpg etc).

Reading RFC9580, "PGP Sign" produces an "ASCII Armour" encoded message, and "PGP Verify" successfully verifies and decodes it - and I think this functionality should be preserved.

Cleartext/Clearsign provides a simpler human-readable message (and the RFC does note some limitations of this format especially relating to reversability).

I think the correct option would be one of:

1. Add an option to the "PGP Sign" operation to produce a Cleartext rather than Armoured output
2. add a new "PGP ClearSign" operation that produces the Cleartext output.

Personally I am leaning towards option 2

In either case the documentation for the "PGP Sign" operation should be amended to make clear what the expected output is.

[engrach]

I'm also leaning to option 2.

Is the way forward to create a new operation that implements this?

[GCHQDeveloper581]

I think that's the way to go, with the addition of some tweaks to the "PGP Sign" documentation (probably new "PGP Clearsign" docs too) to make it clear which implements which output, and where to find the other.

[engrach]

Want me to pickup the PGP Sign code and create the new PGP ClearSign operation from that?

[GCHQDeveloper581]

Sounds good!

Reading RFC9580, and trying some of the Test Vectors in Appendix A, it may be time for a major refresh of the PGP implementation within CyberChef anyway as kbpgp doesn't appear to support the new v6 key standard.

[GCHQDeveloper581]

Incidentally, ref ClearSign:

https://gnupg.org/blog/20251226-cleartext-signatures.html