External Payload Support for CoseSign1

[alex-richards]

What is the areas you would like to add the new feature to?

Go-COSE Library

Is your feature request related to a problem?

It's currently not possible to verify the signature of an external payload.

What solution do you propose?

Add a Verify that takes the payload as a parameter.

What alternatives have you considered?

Setting Payload on the CoseSign1 object . Not great, would refer not to modify the original object.
Constructing SigStructure validating with the key directly. SigStructure isn't exported.

Any additional context?

No response

[shizhMSFT]

It's currently not possible to verify the signature of an external payload.

It is possible as you have considered the alternative

Setting Payload on the CoseSign1 object . Not great, would refer not to modify the original object.

Well, it is not great to modify the original object. However, you can modify the cloned object.

And this is by design as it is called "attach".

[shizhMSFT]

Consider the following example

func VerifyP256(publicKey crypto.PublicKey, external, sig []byte) error {
    // create a verifier from a trusted private key
    verifier, err := cose.NewVerifier(cose.AlgorithmES256, publicKey)
    if err != nil {
        return err
    }

    // create a sign message from a raw COSE_Sign1 payload
    var msg cose.Sign1Message
    if err = msg.UnmarshalCBOR(sig); err != nil {
        return err
    }
    msg.Payload = external // attach the detached content
    return msg.Verify(nil, verifier)
}

If the msg is not meant to be modified, you may consider

func VerifyP256(publicKey crypto.PublicKey, msg *cose.Sign1Message, external []byte) error {
    // create a verifier from a trusted private key
    verifier, err := cose.NewVerifier(cose.AlgorithmES256, publicKey)
    if err != nil {
        return err
    }

    // clone the sign message to attach the detached content
    msgToBeVerified := *msg
    msgToBeVerified.Payload = external
    return msgToBeVerified.Verify(nil, verifier)
}

[shizhMSFT]

Maybe all we need is to improve the doc like adding more examples?

[alex-richards]

Is "attach" documented, it doesn't appear in 1851 at all?

[shizhMSFT]

Is "attach" documented, it doesn't appear in 1851 at all?

No, it is not documented in neither README.md nor the godoc. Do you want to write documentation? or I can send out a PR for them.

[shizhMSFT]

Nonetheless, I think we still need your PR #197 for the implementation of the following utilities

func Sign1Detached(rand io.Reader, signer Signer, headers Headers, payload []byte, ...) ([]byte, error)
func Sign1UntaggedDetached(rand io.Reader, signer Signer, headers Headers, payload []byte, ...) ([]byte, error)

[shizhMSFT]

Is "attach" documented, it doesn't appear in 1851 at all?

No, it is not documented in neither README.md nor the godoc. Do you want to write documentation? or I can send out a PR for them.

I've created a PR #205 to add examples for detached payload in godoc.

[SteveLasker]

@alex-richards, can you weigh in on #205 to see if that unblocks #197?

[alex-richards]

@shizhMSFT Am I correct in thinking that you want to see all the Detached functions removed?

[shizhMSFT]

@shizhMSFT Am I correct in thinking that you want to see all the Detached functions removed?

@alex-richards Since I've written the missing docs in #205, I'd suggest removing all the detached functions except the below 2 utility functions in #197.

func Sign1Detached(rand io.Reader, signer Signer, headers Headers, detached, external []byte) ([]byte, error)
func Sign1UntaggedDetached(rand io.Reader, signer Signer, headers Headers, detached, external []byte) ([]byte, error)

[SteveLasker]

@alex-richards, @shizhMSFT, gentle ping on Issue #195 and PR #197

[alex-richards]

@SteveLasker @shizhMSFT Not a decision I can make.

I need a concrete direction.

Cheers, Alex

[SteveLasker]

Thanks, @alex-richards,
Do you feel comfortable with Shiwei's suggestion?

[SteveLasker]

@alex-richards, if you have any concerns, please do let us know. This is a bidirectional conversation. Our goal is to complete your PR for #197 and then support #198

[alex-richards]

Hi @SteveLasker, my main concern is that this totally changes the intent of the PR. My goal was to add support for verifying detached payloads in a way that didn't introduce a load of fiddling with the internal members of the COSESign[1] structs.

@shizhMSFT's suggestion is to remove all of that, and replace it with a pair of utilities for creating SOSESign[1]s with detached payloads.

The question now is are the detached methods valuable or not? I've only seen one opinion on this point.

[SteveLasker]

@henkbirkholz, @qmuntal, @roywill, @setrofim, @shizhMSFT, @SimonFrost-Arm, @thomas-fossati, @yogeshbdeshpande
Can we get more input on @shizhMSFT proposal

[OR13]

I do not think internal interfaces matter, except in maintainability cost.

What does matter is changing the external interfaces, and we do want to minimize those, so that the library is easy to use.

As I understand #195 (comment)

The suggestion is to export 2 new detached sign convenience functions, and use the attached sign existing functions to satisfy them, reducing overall code.

I am in favor of this proposal.

[thomas-fossati]

I am ambivalent, with a slight preference for @alex-richards's approach, which I reckon has pretty good ergonomics. I understand @shizhMSFT argument about code minimisation, but at the same time, I wonder if that's a false economy in that it does little to increase the usability of the API in "detached" scenarios. My two cents, of course.

[shizhMSFT]

What does matter is changing the external interfaces, and we do want to minimize those, so that the library is easy to use.

From security point of view, minimizing external interfaces minimizes the attack surface so that the library has less chance to have vulnerabilities.

[SteveLasker]

Hi folks,
I'd really like to unblock @alex-richards here. Can we please focus on completing these prs:

• feat: support detached payloads in COSESign and COSESign1 #197
• feat: COSEMac #198