Restricted budgets are something many of us are living 
with, but they're no excuse for trying to build your own 
client authentication from scratch.  It's a major wheel 
that doesn't need reinventing.

David Gillett


> -----Original Message-----
> From: Tiago Filipe Dias [mailto:[EMAIL PROTECTED]
> Sent: July 21, 2003 09:17
> To: [EMAIL PROTECTED]
> Subject: Re: Wi-Fi User Authentication
> 
> 
> Why simply not use FreeRADIUS or even inspite expensive, radiator ? 
> 
> One of the solutions could be radius configurated 
> communicating with a ldap.
> 
> sincero
> 
> On Sun, 20 Jul 2003 13:29:32 -0400
> N407ER <[EMAIL PROTECTED]> wrote:
> 
> > Hi, folks,
> > 
> > We're (I use the anonymous "we" here with apologies) in the 
> process of 
> > setting up a Wi-Fi access point here. Bear in mind that we 
> have little 
> > control over client configuration or consistency--personal 
> computers 
> > would be used, with any OS--and don't want to spend a lot of time 
> > providing technical support.
> > 
> > One of the other groups here went with a product called ReefEdge to 
> > provide Wi-Fi authentication to prevent unauthorized usage; 
> as far as I 
> > can tell from chatting with them, it does pretty much the 
> same as what 
> > we were thinking; however, due to cost, we'd prefer to 
> develop something 
> > in-house or use something open source.
> > 
> > So the plan I had was this:
> > 
> > Set up the gateway with a firewall which would by default 
> redirect all 
> > outgoing tcp/80 traffic to some the local machine, which 
> would have a 
> > "sign-in" page. Users authenticate with their 
> username/password, and a 
> > ruleset is temporarily added to the firewall allowing them 
> full outgoing 
> > traffic. When they are done, they log out, deleting the 
> ruleset (or we 
> > time out their connection after a certain amount of inactivity).
> > 
> > The real question I have is, even if we were to use MAC 
> address matching 
> > instead of IP (iptables has an option in the 2.4 kernel for MAC 
> > matching, as I recall) anyone can grab all the information 
> he needs to 
> > spoof a valid connection from a single captured packet. 
> Now, assuming we 
> > close or timeout connections when the user logs out, he'd 
> have to take 
> > over a connection still in use. There is no guarantee, 
> though, that the 
> > victim client would even notice (nor would we), especially if it is 
> > running something like ZoneAlarm and simply drops, with no 
> ICMP reject, 
> > all unexpected packets. This would mean the attacker could 
> simply pick 
> > up all the responses to his spoofed connections without the victim 
> > noticing.
> > 
> > So how can you prevent this without using something which 
> would require 
> > client-side support, like VPN? VPN is not much of an option 
> for us, I've 
> > been told that a Mac VPN client costs money, and 
> regardless, we don't 
> > want to have to support user configuration. Do I have to 
> simply hope no 
> > one will be able to hijack a connection which is in use?
> > 
> > I've seen software which claims to detect attempts to hijack Wi-Fi 
> > networks, but most appear to just detect brute-forcing on the IP 
> > address, etc. Any attacker could merely passively capture a single 
> > packet and bypass this detection in a snap.
> > 
> > Thanks for any help.
> > 
> > 
> > 
> --------------------------------------------------------------
> -------------
> > Evaluating SSL VPNs' Consider NEOTERIS, chosen as leader by 
> top analysts!
> > The Gartner Group just put Neoteris in the top of its Magic 
> Quadrant,
> > while InStat has confirmed Neoteris as the leader in marketshare.
> >      
> > Find out why, and see how you can get plug-n-play secure 
> remote access in
> > about an hour, with no client, server changes, or ongoing 
> maintenance.
> >           
> > Visit us at: http://www.neoteris.com/promos/sf-6-9.htm
> > 
> --------------------------------------------------------------
> --------------
> > 
> 
> 
> 
> --------------------------------------------------------------
> -------------
> --------------------------------------------------------------
> --------------
> 

---------------------------------------------------------------------------
----------------------------------------------------------------------------

Reply via email to