You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by Jeff Chan <je...@surbl.org> on 2004/06/04 17:37:23 UTC

large number of empty URIs preceeding actual one

Not sure if this is a new type of spam or not:

  http://www.surbl.org/fitch7826drug.us.4jun04.txt

This example I just received had many real or joe job URIs
with no text in the anchor like:

  <a href=3D"http://www.elysian-MUNGED.com"></a>

Perhaps it's trying to run out some counters, but the real
target domain is visible as the last "removal" URI:

  <a href=3D"http://=
  www.ozone.fitch7826drug-MUNGED.us/d.ddd">here.</a>

> Name:    fitch7826drug.us
> Address:  61.250.93.214

Where this IP is in sbl.spamhaus.org of course.

The "ordering" link just before it was broken (no dot, at
least in my MUA, The Bat!):

  <a href=3D"http://fitch7826drug=
  us/b94">Click

Interestingly SpamCop did parse the message correctly in terms
of ignoring the blank anchors and finding only the clickable
ones.

That said, if urirhsbl or SpamCopURI limit the number of
URIs checked, these could sneak through.  A useful behavior
might be to ignore any non-clickable anchors, if we're not
already doing that.

Jeff C.
-- 
Jeff Chan
mailto:jeffc@surbl.org
http://www.surbl.org/


Re: large number of empty URIs preceeding actual one

Posted by John Fawcett <jo...@michaweb.net>.
From: "Jeff Chan" 
> Not sure if this is a new type of spam or not:
> 
>   http://www.surbl.org/fitch7826drug.us.4jun04.txt

Did this get flagged by SpamCopURI? 

If not, the reason is likely to be that the URI

http://=
www.ozone.fitch7826drug-MUNGED.us/d.ddd

contains a quoted-printable character (=).


> That said, if urirhsbl or SpamCopURI limit the number of
> URIs checked, these could sneak through.  A useful behavior
> might be to ignore any non-clickable anchors, if we're not
> already doing that.

I don't think there is *currently* any limit in SpamCopURI
on number of URIs. It should parse all of the URIs. 

Limiting to URIs which are clickable could be a useful 
improvement though to counter spammers that add lots of 
legitimate but non clickable URIs to email. This would 
reduce resource usage during processing and also 
reduce load on surbl servers.

John

Re: large number of empty URIs preceeding actual one

Posted by "Carl R. Friend" <cr...@rcn.com>.
   On Fri, 4 Jun 2004, Jeff Chan wrote:

> Not sure if this is a new type of spam or not:
> 
>   http://www.surbl.org/fitch7826drug.us.4jun04.txt
> 
> This example I just received had many real or joe job URIs
> with no text in the anchor like:
> 
>   <a href=3D"http://www.elysian-MUNGED.com"></a>

   This has been going on for some time now and is designed
to (a) confuse URIRBLs and (b) possibly poison URIRBLs if
they're using highly-automated techniques for URI injection.
They also break up, with legitimate (but useless) HTML syntax
normal words in an attempt to confuse filters.

   Trying to confuse URIRBLs is understandable behaviour for
spammers.  Actively trying to poison them is reprehensible.

   Here's a custom rule I use to catch them:

rawbody CRF_NULL_URL    /<a .{0,16}href=.{0,32}><\/a>/i
describe CRF_NULL_URL   Useless (invisible) HTML link
score CRF_NULL_URL      1.0

   Someone's going to have to look into the URIRBL plug-in
for SA to see if it ignores URIs nested in such constructs
(It should, I believe).

> Perhaps it's trying to run out some counters, but the real
> target domain is visible as the last "removal" URI:

   Since the anchor has no length, it's both invisible and
unselectable; it never gets referenced from the message.

> The "ordering" link just before it was broken (no dot, at
> least in my MUA, The Bat!):
> 
>   <a href=3D"http://fitch7826drug=
>   us/b94">Click

   The spammer didn't know how to use his ratware.

> Interestingly SpamCop did parse the message correctly in terms
> of ignoring the blank anchors and finding only the clickable
> ones.

   That needs verification.

> That said, if urirhsbl or SpamCopURI limit the number of
> URIs checked, these could sneak through.  A useful behavior
> might be to ignore any non-clickable anchors, if we're not
> already doing that.

   What I said.

+------------------------------------------------+---------------------+
| Carl Richard Friend (UNIX Sysadmin)            | West Boylston       |
| Minicomputer Collector / Enthusiast            | Massachusetts, USA  |
| mailto:crfriend@rcn.com                        +---------------------+
| http://users.rcn.com/crfriend/museum           | ICBM: 42:22N 71:47W |
+------------------------------------------------+---------------------+