[linux-elitists] Fwd: PGP signature attachments!

Sean Neakums sneakums@zork.net
Fri Sep 7 14:41:40 PDT 2001

begin  Karsten M Self quotation:

>> There is big difference betwen interpreting a de-facto standard
>> such as "-----BEGIN PGP SIGNED MESSAGE-----" as denoting a signed
>> message and writing a mail client that chokes on what a bad
>> programmer assumed would always be a uuencoded inclusion.  Unlike
>> Outlook, Gnus's interpretation of such plain-text markers can be
>> disabled[0].
> ...and if it's not disabled?  I presume something like the following...
> Version: GnuPG v1.0.6 (GNU/Linux)
> ...might_prove_to_be_unreadable_by_default_within_your_mailer_of
> choice.
> If that doesn't work, there's probably some way to misuse this
> feature which will emerge on further exploration.  That's not my
> aim, it's merely a risk.

Didn't work at all.  I got the usual:

  [[PGP Signed Part:Undecided]]

which gave me

  [[PGP Signed Part:Karsten M. Self <kmself@ix.netcom.com>
  Untrusted, Fingerprint: F932 8B25 5FDD 2528 D595 DC61 3847 889F 55F2 B9B0]]

when I hit RET on it.

Everything can be broken some way; MIME and RFC2015 are so much more
complex than a simple inline PGP message that I really doubt that
either is more or less likely to be exploitable.  But like you, I have
no proof to back *my* statement up.

> Moreover, RFC 2015 includes directives to mail handling utilities
> regarding integrity of messages, and how they are or aren't to
> modify a message text which has been signed or encrypted.  As you've
> certainly read my rant closely by now, you'll note the specific
> reference I've made to the munging issue.  Cleartext signing
> provides no such hints, and there is no assurance your cleartext
> signed message will be delivered intact.

People fail to check signed messages, and it's the MUAs fault for
allowing them to be munged in the first place?  These "hints" may
prevent mistakes, but they will certainly not get in the way of buggy
software or an active attacker.

> MIME is an established and official IETF standard.  RFC 2015 is not
> officially recognized, due to its draft status, but it's a fairly
> widely implemented standard.  The Gnus feature is, by contrast, an
> exploitation of a convention.

Given its non-ratified status, RFC2015 is merely an
extensively-documented convention.

"The man who laughs at standards--that man must be put down.
 We are none of us perfect; I know that.  But we must agree
 on what perfection is."
	-- Joe Gendreau, California Weights and Measures

More information about the linux-elitists mailing list