[OpenWireless Tech] ANYFI IS PROPRIETARY!

[Björn Smedman]

On Wed, Aug 14, 2013 at 4:16 PM, Todd Freeman <todd at chiwifi.net> wrote:
> WPA/WPA2 is good for wifi end points because the attacks against it would
> require a local user, thus it's mostly a non issue that it is easy to break simply
> because of how pointless it usually is. But to use that same method to cross the
> internet ? where it passes though many MANY points where MITM can originate.
> It was never meant to do what you are doing with it, because of this scenario. It's
> good enough security for a hotspot, it is not good enough to secure your traffic
> that is traversing tapped networks. [snip]

I think there is a fair objection in this general direction, basically
that the Internet is a much bigger and scarier place than your average
Wi-Fi coverage area, and that the potential consequences of a bad WPA
passphrase (one that can be broken with a dictionary attack) are
therefore greater. I agree, and we're working on ways to mitigate
that. The longer-term solution however is probably EAP-PWD [1] which
is already starting to gain some device support.

I also agree that MITM should be a primary concern. But I think you're
wrong to assume that our "WPA over the wire" approach would increase
the risk of MITM; in most cases it radically lowers it! In fact that's
one of the major motivations behind the Anyfi.net architecture [2].

This is probably the least understood aspect of IEEE 802.11 security
IMHO: Your protection against MITM hinges entirely on the attacker not
being able to access the authentication credentials/interface [3].
Give me access to your RADIUS authentication interface, and I'll show
you a rogue AP! With that rogue AP I can MITM attack any end-user
device that trusts your network (i.e. has previously connected to it),
anywhere in the world, at any time... And with every new legitimate AP
you install (and configure with RADIUS shared secret, radsec, ipsec,
whatever) you're adding one more single point of failure to your
network from a security perspective.

But it also works the other way around: Make sure that an attacker
cannot access the authentication credentials/interface and you've
pretty much eliminated the risk of rogue APs [4].

Now this is where Anyfi.net really shines; it lets you physically
protect the authentication interface/credentials. The AP doesn't need
access to them because it just forwards raw encrypted IEEE 802.11
frames over an IP tunnel, to a physically secure location where they
can be authenticated without exposing the authentication
interface/credentials to the outside world. This *greatly* strengthens
the system against MITM.

Björn

1. http://tools.ietf.org/html/rfc5931

2. http://anyfi.net/documentation#architecture

3. RFC3579 - RADIUS Support for EAP, 4.3.9. Separation of
Authenticator and Authentication Server,
http://tools.ietf.org/html/rfc3579#section-4.3.9

4. The only way to operate a rogue AP without access to the
authentication interface/credentials is to attack the mutual
authentication of the IEEE 802.11 security mechanism itself, and
that's really only known to be feasible for PSK with a poorly chosen
passphrase and some poorly configured/implemented EAP protocols.