Re: [Secdispatch] [dispatch] [art] Plain text JSON digital signatures

On 2021-04-27 18:21, Carsten Bormann wrote:
> On 2021-04-27, at 17:47, Brian Rosen <br@brianrosen.net> wrote:
>>
>> There was a lot of opposition to the idea previously,
> 
> Yes.

Dear Carsten,
There were indeed a lot of opposition at IETF-104 but nobody has to date bothered providing a single example showing why this idea {w|sh}ould not work [1].

If we take the better [and fully standardized] alternative (JWS), it transforms a JSON object into a Base64Url-encoded string and then puts it into a specific signature container.  That is, the result bears very little similarity to the original which obviously is a drawback. Then a [moderately] skillful attacker replaces the algorithm in the JWS header with the standardized "none" and the whole thing passes validation with flying colors [2].

Anyway, variants of detached (enveloped) JWS signatures in JSON are likely to become a de-facto standard.  Here is an example from a very active group withing the W3C:
https://w3c-ccg.github.io/ld-proofs/#example-2-a-simple-signed-linked-data-document

Feel free rearchitecting https://fido-web-pay.github.io/ using the current JOSE stack; it might even be fun :)

BTW, I have just started the design of a CBOR library needed for dealing with CTAP2/FIDO for the project above.  CBOR seems pretty cool.

Regards,
Anders

1] In all fairness, it does require a bit of work for the application developer who may have to adjust the parsing scheme (not the parser) for things like RFC3339 data.

2] In a n00b world, where developers do not understand that a compliant JWS library does not necessarily come with suitable default policies.

> 
> But there is also some opposition to the weird way this is presented:
> 
>>> On Apr 27, 2021, at 11:27 AM, Bret Jordan <jordan.ietf@gmail.com> wrote:
>>> JWS/CT enables JSON objects to remain in the JSON format after being signed (aka "Clear Text" signing).
> 
> We have a lot of ways that enable signed objects to remain in the format in which they were at signature time.
> 
> Maybe we can fix the presentation of the idea more towards “we really liked XMLDsig and want it back for JSON”, which is certainly a position one can take.
> 
> Grüße, Carsten
> 
> _______________________________________________
> dispatch mailing list
> dispatch@ietf.org
> https://www.ietf.org/mailman/listinfo/dispatch
> 