[repost from earlier posting to spamassassin-devel, where it hasn't shown up
 after 7 hours or so. excuse my ignorance: but is that list moderated?
 http://lists.sourceforge.net/lists/listinfo/spamassassin-devel is not
 hinting to this]
 
Hello,

 I wish to propose the following new method to be implemented in future
versions of SA - but I unfortunately lack familiarity with the code
base, so I am unlikely to do the programming :)

I've had this concept in my head for a couple of months, but have not
seen anyone else uttering it.

1) virtually all spam contains either a URL or an email address, or
   both as a means of contacting the spammers. These are easily
   available with current code for URL/URI rules.

2) hostnames in URLs and email addresses tend to resolve to valid IP
   numbers, and have name service from DNS servers at known and
   valid IP addresses to be functional.

3) people are attempting to use SA for assigning scores based on
   appearance of arbitrary IP numbers (in headers) and domain
   names, as witnessed by
   http://www.stearns.org/sa-blacklist/sa-blacklist.current and
   http://www.merchantsoverseas.com/wwwroot/gorilla/evilrules.cf
   This does not scale: maintaining these rulesets by hand and
   distributing them to a larger audience is exceedingly hard and
   completely impractical.
   
4) IP numbers are relatively easily mapped to geographical regions
   (by RIR : ARIN, LACNIC, RIPE, APNIC, JPNIC) - this is a backup
   classification criteria in addition to 5) :

5) IP numbers are listed in great quantities in DNSBLs, with listing
   criteria as diverse as country/region, open relay/proxy, spam
   source, or spam support services like spamvertized web page
   or DNS hosting. Operation of DNSBLs is a well-established
   'science', it scales, it is manageable, there is plenty of
   choice.

6) 'roaming' websites have appeared that are hosted via reverse
   proxies on 1000's of compromised, trojaned and unfirewalled
   (Windows) machines, for both port 80/tcp (http) traffic, as
   well as 53/udp DNS traffic - with delegated nameservers changing
   records for these sites every few minutes, and keeping extremely
   short TTLs (less than 10 min.)

I propose the inclusion of code and rulesets to achieve the following
three goals, with a fourth one being designated a 'far future' goal:

a) based on the concept of the current DNSBL lookups for IP numbers in
   mail headers: extend that concept to every hostname contained
   in URLs or email address FQDNs found in the message body or any
   header line: Subject:,From: and Return-Path: comes to mind, primarily.

b) based on the concept of a), lookup all host nameserver records (IP
   numbers) for said host/domain names (based on statement #2 above)
   in DNSBL's as well, and permit rulesets to assign scores.
   A rule computing a score based on the NUMBER of such NS records
   (in case some crafty spammer tries to DoS this concept by listing
   200+ DNS servers for his domains), and a limit for the number of
   DNSBL lookups so done to a reasonable number is required.

c) lookup the zone SOA values of a given website's domain name records,
   and assign a score based on arbitrary ranges of these values:
   refresh, retry, expiry, minimum time.

d) future concept: follow the spamvertised URL and determine if the
   page gets redirected to some target page and server that can again
   be treated with goal b)
   

Desired result:
- we can now assign arbitrary scores for spamvertized websites and
  their DNS servers that have their IP addresses appear in any DNSBL,
  or have suspiciously 'mobile' DNS configurations.

Example:
- Just ONE rule assigning a substantial score for every hostname
  resolving to an IP number listed in the cn-kr.blackholes.us DNSBL
  would be enough to reliably cut off whatever air supply Alan Ralsky
  thinks he currently has: Web *AND* DNS-hosting in China, and
  criminal spamvertizing my means of breaking and entering through
  100,000's of open proxies. Take a very deep breath before going
  under, Alan, I say.


Thanks,
bye,Kai


--
"Just say No" to Spam                                     Kai Schlichting
New York, Palo Alto, You name it             Sophisticated Technical Peon
Kai's SpamShield <tm> is FREE!                  http://www.SpamShield.org
|                                                                       |
LeasedLines-FrameRelay-IPLs-ISDN-PPP-Cisco-Consulting-VoiceFax-Data-Muxes
WorldWideWebAnything-Intranets-NetAdmin-UnixAdmin-Security-ReallyHardMath



-------------------------------------------------------
This SF.net email is sponsored by: The SF.net Donation Program.
Do you like what SourceForge.net is doing for the Open
Source Community?  Make a contribution, and help us add new
features and functionality. Click here: http://sourceforge.net/donate/
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk

Reply via email to