[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20120227105530.GA12285@suse.de>
Date: Mon, 27 Feb 2012 11:55:30 +0100
From: Sebastian Krahmer <krahmer@...e.de>
To: oss-security@...ts.openwall.com
Subject: Re: Attack on badly configured Netfilter-based firewalls
Hi,
I know that the 127.0.0.1 trick worked in past, but for loopback
addresses this isnt working anymore since quite a while.
You will get a 'martian destination', regardless of routing
or rp_filter's set. If we talk about a Linux kernel:
ip_route_input_slow()
{
[...]
if (ipv4_is_lbcast(daddr) || ipv4_is_zeronet(daddr) ||
ipv4_is_loopback(daddr))
goto martian_destination;
[...]
}
Or I am doing something seriously wrong. No idea what Solaris
or BSD's are doing.
For 'real' NIC's this trick is however still working, even if
the machine is a host (not a router). This leaves some room for
accessing internal admin interfaces from outside. :)
However, playing with source addresses to defeat firewalls should be
difficult, since most dists enable rp_filter.
my 2ct's
Sebastian
On Mon, Feb 27, 2012 at 01:53:29AM +0400, Solar Designer wrote:
> On Sun, Feb 26, 2012 at 10:05:55PM +0100, Eric Leblond wrote:
> > On Sun, 2012-02-26 at 12:17 -0700, Kurt Seifried wrote:
> > > Are there any helpers that can be abused to open holes in the firewall
> > > externally, or is it only internal clients that can cause problems and
> > > trigger the firewall to improperly allow network traffic in/out.
> >
> > No, attacker has to be on a network directly connected to the firewall.
>
> I guess by "internal clients" Kurt was referring to machines behind the
> firewall (e.g., someone clicking an URL that has a string looking like
> an FTP command embedded in it, thereby triggering the FTP helper to open
> a hole - stuff that was discussed in late 1990s and partially mitigated
> by hardening the helpers at the time), whereas by "attacker on a network
> directly connected to the firewall" Eric means that the attacker may be
> _outside_ the firewall (behind its WAN interface), but on the same
> network segment (e.g., the attacker might have compromised a nearby
> server, such as of another customer at a colocation facility).
>
> It is known that a machine will generally receive and process a packet
> routed to one of its NICs by MAC address even if the destination IP
> address is that of another NIC or even loopback (e.g., it is possible to
> access services bound to 127.0.0.1 in this way - but only from directly
> connected machines). Without rp_filter or equivalent, it is possible to
> have these packets' source addresses match the other NIC's network
> segment. My _guess_ (based solely on the info posted in here so far) is
> that the gist of Eric et al.'s new attack is to apply this approach
> against a protocol helper. The novelty is thus in combining these known
> things together to arrive at something that to the best of my knowledge
> has not yet been discussed.
>
> I suppose Eric will tell us if this is the correct guess or not. ;-)
>
> Alexander
--
~ perl self.pl
~ $_='print"\$_=\47$_\47;eval"';eval
~ krahmer@...e.de - SuSE Security Team
---
SUSE LINUX Products GmbH,
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer, HRB 16746 (AG Nürnberg)
Maxfeldstraße 5
90409 Nürnberg
Germany
Powered by blists - more mailing lists
Please check out the
Open Source Software Security Wiki, which is counterpart to this
mailing list.
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
