Scripts send encrypted email to Spamcop

Jesse Norell wrote:
> Ok, from Crypt::CBC man page it's noted:
> 
>        The "salt" header is now the default
as of Crypt::CBC version
> 2.17. In all earlier versions "randomiv" was
the default.
> 
> Looking at the process-quarantine-sub.pl script, it did
not specify the
> header type, and I was getting errors from it in cron
saying:
> 
>   "Ciphertext does not begin with a valid header
for 'salt' header mode"
> 
> That's easy to fix (and may be done in more recent maia
versions, I
> don't know) by just adding the 'header' =>
'randomiv' to the
> Crypt::CBC->new() initialization.
> 
> The amavisd-maia executable already had that done. 
With updating a
> couple other settings there to their new names in both
executables, I
> now have:
> 
>     $cipher = Crypt::CBC->new( {'key' => $key,
>                                 'cipher' =>
'Blowfish',
>                                 'literal_key' => 1,
>                                 'padding' => 'null',
>                                 'header' =>
'randomiv'
>                               } );
> 
> I'm guessing that might fix my bayes problem, and
possibly/probably your
> reporting issue, too.  It did not make the gui able to
display the raw
> message though .. still working on that one.  Any ideas
there?

In your case you're running Maia 1.0.0, which breaks with
Crypt::CBC
2.17 and later on two fronts--it breaks compatibility with
Mcrypt
(addressed by ticket #280 in Maia 1.0.1) and it breaks the
decryption
process in process-quarantine-sub.pl (addressed by new
ticket #368).

The #280 issue with Mcrypt affects the ability of the PHP
scripts to
properly decrypt what Crypt::CBC encrypts.  With the changes
in
Crypt::CBC 2.17, the "RandomIV" header was no
longer being prepended to
the encrypted data, and at the PHP end of things the Mcrypt
library was
relying on that "RandomIV" header to know what to
do with the data.  Not
seeing the "RandomIV" header anymore, the GUI had
to assume that the
data was not encrypted, which is why you're seeing garbage
in the mail
viewer.

Adding the 'header' => 'randomiv' option was indeed the
fix for that
issue, as implemented seven months ago in revision 956
<https://secure.renaissoft.com/maia/changeset/956>. 
This fix was
incorporated into Maia 1.0.1.  Note that 'literal_key' =>
1 is the same
as 'regenerate_key' => 0; the latter is just the
deprecated form of the
former, which was introduced in Crypt::CBC 2.15.

The #368 issue is different however.  For some reason that
I'm
investigating, with versions 2.17 and later of Crypt::CBC an
item
encrypted and then decrypted with the same set of options
does not seem
to decrypt, which suggests that a slightly different set of
options may
be necessary at the decryption end now.  This issue affects
only the
process-quarantine-sub.pl script, so both Bayes training and
spam
reporting are broken when Crypt::CBC 2.17 or later are used.
 This is in
spite of the fix from #280.

In short, #280 is responsible for the garbage you see in the
mail
viewer; #368 is responsible for the ineffectiveness of your
Bayes database.

-- 
Robert LeBlanc <rjlrenaissoft.com>
Renaissoft, Inc.
Maia Mailguard <http://www.maiamail
guard.com/>

_______________________________________________
Maia-users mailing list
Maia-usersrenaissoft.com
http://www.renaissoft.com/mailman/listinfo/maia-users