New versions of GnupG and GPGME

[Werner Koch]

Hi,

we have released new version of GnuPG and GPGME due a forthcoming
security advisory.  That advisory was scheduled for today but I have
not yet seen it.  Given that the night has already fallen over Europe
and that I am about to leave the office, I attach our own advisory for
those who are keen to know what is going on.


Salam-Shalom,

   Werner


             Multiple Messages Problem in GnuPG and GPGME
            ==============================================
                              2007-03-05


Summary
=======

Gerardo Richarte from Core Security Technologies identified a problem
when using GnuPG in streaming mode.

The problem is actually a variant of a well known problem in the way
signed material is presented in a MUA.  It is possible to insert
additional text before or after a signed (or signed and encrypted)
OpenPGP message and make the user believe that this additional text is
also covered by the signature.  The Core Security advisory describes
several variants of the attack; they all boil down to the fact that it
might not be possible to identify which part of a message is actually
signed if gpg is not used correctly.

[ Please do not send private mail in response to this message.  The
  mailing list gnupg-devel is the best place to discuss this problem
  (please subscribe first so you don't need moderator approval [1]). ]


Impact
======

All applications using GnuPG without properly using the status
interface to verify signed or signed and encrypted messages.

All GPGME versions up to and including 1.1.3.

Starting with version 1.4.7 and 2.0.3, GnuPG implements an additional
and sufficient protection against this common usage problem.

Detached signatures are in no way affected by this problem.


Description
===========

When using gpg (or gpg2) in a pipeline or with redirected input and
output additional data may be inserted into a message.  This allows to
forge a signed message by prefixing it with arbitrary material.  A way
to create such a message is:

  echo "This is my sneaky plaintext message" > foobar.txt
  gpg -z0 --output prefix.gpg --store foobar.txt
  cat prefix.gpg original-signed-message.gpg > forged.gpg

Using gpg naively this results in:

  $ gpg <forged.gpg
  This is my sneaky plaintext message
  Either I'm dead or my watch has stopped.
                  -- Groucho Marx's last words
  gpg: Signature made Mon Feb 26 09:57:04 2007 CET using DSA key ID 68697734
  gpg: Good signature from "Alfa Test (demo key) <alfa at example.net>"
  [...]

and thus gives the impression that the sneaky message is part of the
signed Groucho quote.  The correct way to use gpg with redirection is
by taking care of the status interface:

  $ gpg --status-fd 1 <forged.gpg
  [GNUPG:] PLAINTEXT 62 1172479053 foobar.txt
  [GNUPG:] PLAINTEXT_LENGTH 36
  This is my sneaky plaintext message
  [GNUPG:] PLAINTEXT 62 1172480224 original-signed-message
  [GNUPG:] PLAINTEXT_LENGTH 86
  Either I'm dead or my watch has stopped.
                  -- Groucho Marx's last words
  gpg: Signature made Mon Feb 26 09:57:04 2007 CET using DSA key ID 68697734
  [GNUPG:] SIG_ID UncMPBJYgbG/uszJVNKoCAz+hvY 2007-02-26 1172480224
  [GNUPG:] GOODSIG 2D727CC768697734 Alfa Test (demo key) <alfa at example.net>
  gpg: Good signature from "Alfa Test (demo key) <alfa at example.net>"
  [...]

Here the PLAINTEXT status lines clearly identify the start of a new
message.

Note, that using gpg on the command line is in almost all cases not
done with redirection but by letting gpg save the the signed message.
In this case gpg will save the message to different files or in case
the file names are identical, prompt the over to overwrite the first
one again.

Because the problem of identifying the actual signed content when
mixing the signed data and the signature is very common, the long
standing suggestion for all digital signatures is to use a detached
signature.  A detached signature allows to clearly identify what is
signed and what is the signature.  This is also the reason why
PGP/MIME signed messages are in general to be preferred over the old
style clear signed messages.


Solution
========

Given that there are many applications in use which are subject to the
described problem, we have decided to change GnuPG so that such forged
OpenPGP messages are detected and the signature verification will
fail.  GnuPG 1.4.7 has been released today and is available from the
usual places [2].  If you don't want to update, a minimal patch
against GnuPG 1.4.6 is available at

 ftp://ftp.gnupg.org/gcrypt/gnupg/patches/gnupg-1.4.6-multiple-message.patch

Many applications are using the library GPGME which implements an easy
way to process OpenPGP messages using gpg.  We have updated GPGME to
make it immune against this problem even if an old version of gpg is
being used.  GPGME 1.1.4 is available from the usual places [2].  A
patch (against version 1.1.3 or 1.1.2) is available at

 ftp://ftp.gnupg.org/gcrypt/gpgme/patches/gpgme-1.1.3-multiple-message.patch

Please note that - after applying one of these patches - some
vulnerable applications (mainly MUAs) may fail to handle certain
messages which are composed of several OpenPGP messages.  To continue
the support of such messages fixing the application is required as
there is no way for GnuPG to do it.


Support 
=======

g10 Code GmbH [3], a Duesseldorf based company owned and headed by
GnuPG's principal author, is currently funding GnuPG development.
Support contracts or other financial backing will greatly help us to
improve the quality of GnuPG.


Thanks
======

Gerardo Richarte found this problem.  David Shaw greatly helped to
analyse and describe the core of the problem.




[1] See http://lists.gnupg.org/mailman/listinfo/gnupg-devel
[2] See http://www.gnupg.org/download/ 
[3] See http://www.gnupg.org/service.html