In <[EMAIL PROTECTED]> "Fergie" <[EMAIL PROTECTED]> writes:

> -- wayne <[EMAIL PROTECTED]> wrote:
>
>>I'm afraid you have almost completely missed the point of Williams
>>"HELO attack".
>
> Sorry to burst everyone's bubble, but DNS has been on the 'Top 20'
> vulnerability report for the past 7 years. get over it.
>
> Seven years in a row:
>
>  http://www.sans.org/top20/?ref=1814

Yeah, but that just mentions DNS reflector attacks and cache
poisoning.

William's attack is different.  All you have to do as an attacker is
publish "valid" records and get people to do normal DNS lookups on it.

Lots of people publish NS records that cause lame delegations.  Lots
of people have broken MX records.  The only difference here is the
quantity of these records and, of course, the intent.


I just checked this morning when all my zones have propagated and
caches have timed out.  A *single* lookup on clinte.schlitt.net
triggers 40 packets to and another 40 packets from a victim domain
(the name servers for example.com, iana-servers.net), and another
40/40 packets involved a gtld-servers.net server.  As the attacker, it
put a 3/3 packet load on my name server.

Checking the sizes of the packets, it looks like I could probably
increase number of NS records in bad.schlitt.net and ugly.schlitt.net
from 20 to maybe 30, and clinte.schlitt.net could easily be involved
with more just "bad" and "ugly" with its NS records.  (Yeah, this NS
record attack uses just two levels of recursion on the attacker side.)



Now, remember above where I mentioned that lame delegations are all
too common?  OK, right now clinte.schlitt.net gives a SERVFAIL, but if
I add just *one* valid NS somewhere in there, it could be used for any
normal purpose that any other domain could be used for.  Put a bunch
of web bugs into HTMLized spam or on free pr0n pages, and *poof*, a
nice DoS attack.  Heck, on the pr0n pages, you would probably put the
real images on the domain name that triggers the DoS attack, that way
you would be sure that the DoS was completely done.


> Doesn't that tell you guys something?

Yeah, it tells me that DNS was developed at a time when people were,
at worst "having fun" in the compsci classes and abusive behavior
could be caught and whacked pretty easily.

Being liberal with what you accept only works if you know that
everyone else is honestly trying to be conservative with what they
send. :-<


When I was involved with the development of SPF, I was concerned
about DoS potentials and worked hard to make sure that limits were put
into it.  I'm not seeing any limits to prevent DoS attacks via DNS.


As Stuart Gathman mentioned on an SPF list, maybe there needs to be a
count of the number of lame delegations or broken MX records and only
allow a few before declaring the whole thing as "filthy" and
stopping.  This kind of check needs to be put into *every* protocol
that does any sort of indirection on DNS records.  This includes all
name servers (NS records), mail systems (MX lookups), anti-spam
systems (verifying the HELO domain, SMTP callback verification, SPF,
URL checks, etc.), anything that uses SRV or NAPTR records, etc.


-wayne
.
dnsop resources:_____________________________________________________
web user interface: http://darkwing.uoregon.edu/~llynch/dnsop.html
mhonarc archive: http://darkwing.uoregon.edu/~llynch/dnsop/index.html

Reply via email to