I already switched to using OpenVPN over an otherwise unencrypted Wi-Fi network a while back. Obviously the unencrypted network is entirely separate from the rest of the LAN and only the OpenVPN server is listening on it. Getting all my hardware to play nice with WPA was getting very fiddly, and it even seems a little faster doing the encryption entirely in software. Plus, I have more faith in the OpenVPN guys than the people who seem to be unable to release a new standard within a couple of years.
It's also a lot easier to segregate traffic this way. Once you're coming in via openvpn, the traffic is on a completely different interface and IP range, making automatic/transparent proxying a breeze and firewall/iptables setup really straight forward.
Absolutely. It definitely beats going with your average access point's default of bridging wifi to your LAN, though it does make your wiring slightly more complicated. I use two separate VLANs actually, one for the unprotected wifi access points and one for the actual LAN traffic. You have to watch out that you use high-quality VLAN-capable switches, however, as some of the cheaper ones fall over and supply the world with your protected packets if you fill up their MAC address table with spoofed packets... which is of course easy to do via an unprotected access point. That kind of thing might stil DoS you with a good switch, but it shouldn't open up your network.
The other good thing about doing it this way is that you can get device _and_ user level authentication (by using client certificates and password authentication). No more lame MAC filtering on your wireless firewall to limit access. The access point becomes just a dumb ethernet <-> wireless bridge.
Realistically, I don't see a way other than strong, SSL-style crypto with private keys and AP certificates. I'd like to see a crypto expert analyse and write about RADIUS-authenticated WPA2 in a way us mortals can understand. My understanding is that it should be safe, but I'd like to hear about it from someone like Schneier. A lot of APs actually support RADIUS, but you do still need to set up your own RADIUS server somewhere on the network. I guess the open-source firmware available for some APs could help here.
My current OpenVPN solution is actually pretty user-friendly - it doesn't "magically" connect as soon as my AP comes into range, but it is just one click to start. It's almost certainly possible to run the script from the event directly. The nice thing is that you can implement this now, and OpenVPN runs on Linux, OSX, Windows, probably the BSDs too.
I also have an OpenVPN server that's accessible from the internet, which is handy to use from public wifi. I realise that it's not end-to-end encryption, but having my cookies etc. available to everyone around is just a little too scary.
With the US legal situation that might be feasible, but in some countries you can get yourself into pretty deep trouble. From what I remember, German law in particular makes this a really bad idea. (aside from the fact that cease-and-desists cost money in Germany) Austrian law is often quite close to the German counterpart so I'm not going to risk it, even though I strongly doubt anyone would try to leech my wifi around here. (mine is the only wifi signal I can detect, and we get about one car per hour on our road, so not exactly where it's at)
EDIT: I'm no lawyer, and it's been a while since I read the articles on the topic, but the issue was effectively that you fall into one of two categories operating an open AP:
- you get treated as an ISP. In that case I think you need to establish some kind of (implied or explicit) contractual relationship with your "customers" (paying or not) and are subject to various regulations. Hardly feasible.
- everything your "users" do is your responsibility.
I misremembered some of this stuff, and there are no doubt various subtleties I didn't go into. The subject was covered in various articles in the popular German c't magazine. http://www.heise.de/ct/
"Partial crack".
"They have not, however, managed to crack the encryption keys used to secure data that goes from the PC to the router in this particular attack."
Is that a trivial next step, or is the headline misleading?
I'm not 100% sure on how WPA works, but I think the crack allows you to log onto the protected network and exchange data with everyone else on the network. However, it doesn't allow you to sniff traffic that isn't headed your way. For example, Alice is using her laptop to load data from Bob's PC, which is attached to the access point. Eve cracks the AP's WPA passphrase and logs onto the network. She can initiate traffic to Alice and Bob's machine, but she can't see the data they're exchanging because individual wifi users' encryption is negotiated separately.
I think this is the same crack, they're just ramping up the hype slowly, as publicity for whatever conference at which the details will be presented. Like lots of the high-profile "OMG h4x" stories this year. :-/
Yup your correct. The WPA crack has been about for a while.
The older WPA algo's can be broken totally given time (mostly to collect a LOT of data - something like 4 to 20 million packets sometimes).
We couple a small laptop used for collections & injection (to generate data) with our cracking cluster back at our office. Given about 3 or 4 days of activity on a WPA network it is usually possible to get in :)
In terms of coWPAtty: we are work on high level corporate networks that are usually immune to that kind of attack.
We use full on brute force rainbow tables with some custom ideas / programs (obv can't share much of that). In it's simplest form we have 8 file servers, 1000 nodes and 2 head servers. 150+TB's of rainbow tables. Basically we combine the shortcut of rainbow tables with intelligent analysis of the packets and the brute force of the cluster.
For WPA we use some of our own techniques to intelligently make use of the data that gets sent back. Usually we can start to work on the data after 1 day of collection and it can be possibly 2 more days before it is cracked (depending on the other work running on the cluster)
Currently we have 2 specialists employed who are working with me to find ways to break the time down further (with great success) and I am also currently testing and cycling up another 9,000 nodes to add into the cluster.
By early next year we hope to be able to turn up at a location and crack WPA networks in a days work :)
In terms of economy it is not great (working on that) as your looking at about £5-8,000 outlay just to break in - but for speed I believe it is unmatched.
Bear in mind ofc this is not representative of what a hacker could do. The rainbow tables (for all our decryption not just WPA) have taken 8yrs to make so far (on around 5000 nodes) and we still have about 2yrs more work to do on them. What this article talks about is a shortcut technique but ultimately I dont beleive it will lead to an inherent weakness which can bypass the need for our type of attack. But what it will be useful for is forcing the network to generate traffic :)
Hopefully sometime next year we should be read to deploy our full AES attack - but that (based on current testing) is going to require a LOT of nodes (possibly pushing 25,0000).
