Karsten Bräckelmann wrote:
> Please see bug 6155 comment 128.
> https://issues.apache.org/SpamAssassin/show_bug.cgi?id=6155#c128
> 
> The issue persists, and being an onet.pl customer using The Bat!
> MUA is kind of a deadly combination. Those are being shot
> single-handedly.
> 
>   score KB_DATE_CONTAINS_TAB   3.800 3.799 3.799 2.751
>   score KB_FAKED_THE_BAT       2.432 3.441 2.008 2.694

Karsten forgot to mention that the latter rule is dependent on the
former rule.

> Comments? Is this something we need to worry about? I do, rather
> scary to me. :/

Me too.  Is onet.pl the only such instance of this overlap?  Why not
move KB_DATE_CONTAINS_TAB to a subrule and make some kind of
onet.pl-detector that the new meta must match against?

They have an SPF record, though we probably can't rely upon
implementations using SPF, so since it's simple enough, we could
manually parse it:

header __DATE_CONTAINS_TAB  Date:raw =~ /^\t/
header __RCVD_ONET_PL  X-Spam-Relays-External =~
/ ip=213\.180\.1(?:28|[345]\d)\.\d/
meta   KB_DATE_CONTAINS_TAB __DATE_CONTAINS_TAB && !__RCVD_ONET_PL
# here for reference, unchanged:
meta   KB_FAKED_THE_BAT     __THEBAT_MUA && KB_DATE_CONTAINS_TAB

Still, this is bad as it assumes their SPF record is foolproof and
unchanging (their use of "-all" shows that it is at least
theoretically authoritative).  Also, since we need it to survive
mailing lists and forwarding services, we can't key an IP check on the
last external relay, meaning that the passage through onet can be
forged.  Given some other headers, we might be able to find a better
counter-filter.


To attack it from another angle:  has anybody (preferably a Polish
speaker) approached Onet.pl and at least mentioned the problem if not
also a solution?  Their use of SPF implies that they care about this
sort of thing...

Reply via email to