Graham Toal wrote:
> We have an internal network in a public area that anyone can walk up to,
> and plug in a PC. 
> I want someone hooking up to the network to have NO access to the outside,
> *until*...
> 
> the first time they use a web browser to access any outside page, it
> is redirected to a browser on the firewall host.  That browser puts up
> a page requesting a username and password which it checks in some
> database it has access to.
> 
> Once the user has been validated, the ip chains are modified to allow
> that host full routed access to the net. (For a specific length of time -
> a timer will kick off and when that time expires, another script will be
> run to remove the rules which permitted that IP access)
> 
> This is basically the same system as some hotels run for internet access
> from your room, except that they ask for a credit card whereas we ask for
> a valid student username and password.  (This is for a university environment)
> 
> Has anyone done this before?  If so please point me at it!

I posed the same kind of question to this list in Aug 2001:
  http://msgs.securepoint.com/cgi-bin/get/netfilter-0108/200.html

That particular project never got traction, but what I learned may be of help.
I'd very much like to keep this discussion going, in list preferably, since
this issue of public-access networking is timely, interesting and applicable
to netfilter community. 

There are commercial products that do these functions but at high cost
($10-15K), notably:
Cisco's BBSM (broadband Building Service Manager) and
SolutionInc (http://www.solutioninc.com/products/hospitality.html)

Being primarily for MxU markets, these solutions all have a billing component
to hook into (for example) the Hotel PMS system. They also support other auth
models like username/password input, not just credit card models. For example,
when someone is renting a conference room and needs to allow all participants
to connect, a pre-assigned username/password model is used.

I believe the commercial offerings use proxy arp to force requests to an
authentication/billing service running locally. Whether they use dynamic fw
rules I can't say but it would seem likely that they do.

What you may/maynot also want is a way to segregate LAN segments and/or
individual users from one another. In the case VLAN/802.1q can be used in
conjunction with switched network which supports VLAN tagging. Recent linux
kernels > 2.4.15 support VLAN/802.1q - 
  http://scry.wanfear.com/~greear/vlan.html
and some notes on how I got VLAN working with IPtables and DHCP:
  http://www.planetconnect.com/vlan/

some tangentially related links included FWIW:
NetReg: An Automated DHCP Registration System 
 http://www.usenix.org/event/lisa99/full_papers/valian/valian_html/index.html

Dealing with Public Ethernet Jacks - Switches, Gateways, and Authentication 
  http://www.usenix.org/event/lisa99/full_papers/beck/beck_html/index.html

Again...This is a great start but I'd like to keep this discussion going so if
anyone has other ideas/info/insight to add, please chime in!
--
Doug Monroe

Reply via email to