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

[Anders Rundgren <[e-mail]>]

On 2021-04-28 11:29, Stefan Santesson wrote:
> RFC 7797 is supported by common open source such as Nimbus and I use it
> for instances where you obviously do not need a URL safe token.

I see.

> As such it works for JWS but Not for JWT. But It works really well and
> saves space when URL safeness is not needed.

That's great.  Note though that this scheme require that '.' are escaped into \u002e as can be seen in the example below.  Data to sign:
{
   "payee": "Space Shop",
   "amount": {
     "currency": "EUR",
      "value": "100.00"
   }
}

Using RFC 7797 (with "b64":false,"crit":["b64"]):
eyJhbGciOiJFUzI1NiIsImI2NCI6ZmFsc2UsImNyaXQiOlsiYjY0Il19.{"amount":{"currency":"EUR","value":"100\u002e00"},"payee":"Space Shop"}.KFOhCfpyZnW7RUpwOvZAJPKDfpF1R2uENirUW6Ew4v1HwQI5iFJaXKW0PsvTuVNpb-T_UjpOcv868qihMjeMwA

Using JWS/CT (here pretty-printed using standard JSON tools):
{
   "payee": "Space Shop",
   "amount": {
     "currency": "EUR",
      "value": "100.00"
   },
   "signature": "eyJhbGciOiJFUzI1NiJ9..aY27xoS5B3J-uZtUdJzXevqBbnNf4vNT1YN1_eHPOcILMXlVu3VjdExW17f66EGdt4mxQSnpAVLsUa4k2zSLQw"
}

> So I guess your answer is that it still encapsulates the signed JSON in
> the signature, and that the proposal really is about embedding signature
> in the JSON object being signed (and not about whether the JSON is in
> plaintext).

JWS/CT addresses several issues:
- Maintaining a consistent message structure
- Maintaining JSON notation
- Full compliance with JavaScript and browser APIs (including the "JSON" object).

However, from an IETF adoption point of view it is rather the validity and reliance on RFC 8785 that is the core issue.

> Could you elaborate why you think it is important to NOT embed signed
> data in the signature?

The example above may serve this purpose.

> What is the usecase?

See this response to Carsten: https://mailarchive.ietf.org/arch/msg/dispatch/3EX1uM9K_oCft7BoQfW3EQj_En4/

thanx,
Anders

> 
> /Stefan
> 
> 
> On 2021-04-28 11:00, Anders Rundgren wrote:
>> On 2021-04-28 8:52, Stefan Santesson wrote:
>>> How is this different/better than implementing RFC 7797 and apply the
>>> header b64=false in order to carry plaintext JSON in the payload?
>>
>> Good question!
>>
>> Apart from the fact that the data becomes embedded in the JWS
>> signature container (=changing the structure), you cannot really put
>> JSON there:
>> https://tools.ietf.org/html/rfc7797#section-5.2
>>
>> My guess that the only real-world use of this option is to save an
>> internal-only (but technically redundant) Base64Url-operation for
>> truly detached data, be it it JSON, PNG, etc.
>>
>> JWS/CT was designed for signing JSON Objects ({}), and let them remain
>> as such.
>>
>> Thanx,
>> Anders
>>