Date: Wed, 24 Mar 2004 14:36:22 +0900 (JST) From: Motonori Shindo <mshindo@mshindo.net> To: garycor@comcast.net Cc: freebsd-net@freebsd.org Subject: Re: WEP problems with ndis and ath drivers Message-ID: <20040324.143622.59463083.mshindo@mshindo.net> In-Reply-To: <406108F7.3030704@comcast.net> References: <20040321013533.GA37342@panzer.kdm.org> <B0667184-7CD1-11D8-AA7B-000A95AD0668@errno.com> <406108F7.3030704@comcast.net>
next in thread | previous in thread | raw e-mail | index | archive | help
Gary,
From: Gary Corcoran <garycor@comcast.net>
Subject: Re: WEP problems with ndis and ath drivers
Date: Tue, 23 Mar 2004 23:05:11 -0500
> Sam Leffler wrote:
>
> > It appears your AP requires shared-key authentication to associate when
> > WEP is enabled. The current code in the tree does not support
> > shared-key authentication (it's actually a bad idea security-wise).
>
> Hence I'm curious why, if "open" equates to "no" authentication,
> you suggest that shared-key authentication is a worse option?
> Perhaps it is - I'm just trying to learn a bit more...
Shared-key authentication is in fact a worse option than open
authentication. Basic idea how shared-key authentication works is as
follows:
Station Access Point
Auth Req .... (1)
----------------------->
Challenge .... (2)
<----------------------
WEP(IV+Key, Challenge) .... (3)
----------------------->
Auth OK .... (4)
<----------------------
Access Point challenges the Station with random number (128
octets). Station then encrypts it using WEP with a key shared by both
Station and Access Point, and send it back to the Access Point. Access
Point validates the reply by first decrypting the packet and then
calculating the ICV. If ICV tells it is OK, then Access Point grants
the access.
Suppose that malicious user sniffs this authentication sequence.
Malicious
Station Access Point
Auth Req
-----------------------> .... (a)
Challenge
<---------------------- .... (b)
?????
-----------------------> .... (c)
Auth OK !!
<---------------------- .... (d)
The first two steps (step (a) and (b)) is just like the legitimate
case (step (1) and (2)). Because malicious user doesn't know the WEP
key, it may look that (s)he has no way to send a correct challenge
response in step (c). In fact, this is exactly where this
authentication scheme is broken!! Because WEP is based on RC4 stream
cipher, XORing the messages in step (2) and (3) recovers the "key
stream" associated with a given IV (note that this is not the WEP
key). Malicious user then computes "(2) XOR (3) XOR (b)" to come up
with a challenge response and sends it back to the Access Point with
IV observed in step (3). This challenge response will be accepted the
Access Point as valid even if malicious user doesn't know the WEP
key!! Well, I intentionally omitted a few minor points
(e.g. authentication frame format, linearity of CRC32 and XOR, etc.)
in this explanation for brevity, but this attack can still be mounted
anyway.
You can easily see that this authentication scheme is in fact worse
than nothing (open authentication). Access Points leaks key stream of
first 128 octets every time this authentication is performed just for
free to everybody (including malicious users).
Considering all this, Access Point should always reject shared-key
authentication even if Station requests it.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040324.143622.59463083.mshindo>