Re: [openpgp] A way to securely define cleartext signature charset

[Andre Heinecke <[e-mail]>]

On Tuesday, September 11, 2018 11:53:54 AM CEST Werner Koch wrote:
> Verification tools already need to consider an unsigned armor header to
> figure out the digest algorithm to use.  However, this is sometimes not
> enough because some tools used to have peculiar interpretation of white
> space and line endings or the "Hash" header line was missing.  Thus, for
> one-pass processing running a second hash context was (or well, is)
> useful.  Adding a "Charset" header and automatically try to convert
> would lead to an even more complex verification step.  I don't think
> that is justified.

Thinking more from the "backend" standpoint and less from the Application 
using the backend this makes sense to me. A minor issue is that my Application 
might temporarily show the wrong representation before the verification is done 
but I guess that is indeed minor.

> Better have a way to sign the character set info and present this to the
> user in the Good and in the Bad verification case.  On a bad
> verification the user can then choose to convert and try a verification
> again.  That would not be a one-pass processing anymore but for the ugly
> cleartext signatures this seems to be acceptable.

Yes, as for me the "User" would be my Application and not the person sitting 
in front of the Computer I think that is acceptable as it can be handled 
automatically.

> I would thus suggest this new standard notation:
> 
>   ##### The 'charset' Notation
>   
>   The "charset" notation is a description of the character set used to
>   encode the signed plaintext.  The default value is "UTF-8".  If used,
>   the value MUST be encoded as human readable and MUST be present in the
>   hashed subpacket section of the signature.  This notation is useful
>   for cleartext signatures in cases where it is not possible to encode
>   the text in UTF-8.  By having the used character set a part of the
>   signed data, attacks exploiting different representation of code
>   points will be mitigated.

I like it.

"The default value is "UTF-8"" -> Do I understand this correctly that this 
basically means: If no charset notation is provided a cleartext signature MUST 
be in UTF-8?
That would be great.


Thanks and best regards,
Andre

-- 
Andre Heinecke |  ++49-541-335083-262  | http://www.intevation.de/
Intevation GmbH, Neuer Graben 17, 49074 Osnabrück | AG Osnabrück, HR B 18998
Geschäftsführer: Frank Koormann, Bernhard Reiter, Dr. Jan-Oliver Wagner