As far as I can tell, all of these attacks require the attacker to already be associated to a victim's network. Most of these attacks seem similar to ones expected on shared wifi (airports, cafes) that have been known about for a while. The novel attacks seem to exploit weaknesses in particular router implementations that didn't actually segregate traffic between guest and normal networks.
I'm curious if I missed something because that doesn't sound like it allows the worst kind of attacks, e.g. drive-by with no ability to associate to APs without cracking keys.
The attacker doesn't need to be connected to the victim's network, only to the same hardware, the hardware's loss of isolation is the unexpected problem.
Their University example is pertinent. The victim is an Eduroam user, and the attacker never has any Eduroam credentials, but the same WiFi hardware is serving both eduroam and the local guest provision which will be pretty bare bones, so the attacker uses the means described to start getting packets meant for that Eduroam user.
If you only have a single appropriately authenticated WiFi network then the loss of isolation doesn't matter, in the same way that a Sandbox escape in your web browser doesn't matter if you only visit a single trusted web site...
I should reinforce this point by saying that it's the default position for "guest" networks to be using the same hardware as "secure" office wifi and such.
I'd further reinforce this by pointing out that this is what the specific term, guest network, means - it's the common name used by router manufacturers to describe an optional feature of serving secondary network from the same hardware, intended for the specific, common use case of serving transient and/or less trusted users.
This is in contrast to more genetic, descriptive terms like "additional network", "separate network for guests", etc.
Yeah, that commercial-grade hardware didn't actually isolate at the PHY-MAC layer is a bit surprising. How would they have working VLANs at the AP?
802.11 is kinda poorly designed in this regard, but they do isolate to some degree. I need to read the paper, some claims here have a very strong "misunderstood or wrong or specific vendor problem" smell.
What about XFinity, which by default shares the wifi you pay for with strangers to create access points around the city?
It sounds like this attack would work in that scenario provided the attacker is able to connect to the guest access point.
I haven’t paid attention to one in a while but I seem to remember the need to authenticate with the guest network using Xfinity credentials. This at least makes it so attribution might be possible.
It looks like both clients must be on the same VLAN for the attack to work. They could be connected on different BSSIDs or even different SSIDs, but they still must be on the same VLAN.
If the vulnerability is between layers 1 and 2, wouldn’t that imply that VLAN tagging at layer 2 might not be effective in segregating the traffic?
Wireless cards typically don't expose the VLAN tags directly. So VLANs should be OK.
This is probably the biggest issue.
I turn WiFi mine off and use my own WiFi ap.
Yeah, along these lines I've always been biased strongly against using ISP hardware beyond the minimum required to connect to the outside world.
See also: Amazon's Sidewalk (which shares your network via Ring camerae, e.g.).
[dead]
I'm a co-author on the paper: I would personally indeed not use the phrase "we can break Wi-Fi encryption", because that might be misinterpreated that we can break any Wi-Fi network.
What we can do is that, when an adversary is connected to a co-located open network, or is a malicious insider, they can attack other clients. More technically, that we can bypass client isolation. We encountered one interesting case where the open Wi-Fi network of a university enabled us to intercept all traffic of co-located networks, including the private Enterprise SSID.
In this sense, the work doesn't break encryption. We bypass encryption.
If you don't rely on client/network isolation, you are safe. More importantly, if you have a router broadcasting a single SSID that only you use, we can't break it.
Hi and thanks so much for the valuable research!! I know it has been asked a lot here already, and probably some in-deep reading would help figure that out by myself. But I’ve noticed that you used Cisco 9130 APs, and noticed only part of the attack work on those. So wanted to ask whether you tested those with just IP based network separation, or also the VLAN-based one? Also, since you’ve mentioned the findings have been communicated to the vendors and the WiFi alliance alike, may I ask you to maybe share a CVE number here? I (as probably a lot of us here), use some of the hardware mentioned for personal goals/hobby in my home setup, and find it fun to keep that setup reasonably protected for the sake (fun) of it. Much appreciated!
We don't have a CVE number. Whether devices/networks are affected also highly depends on the specific configuration of the device/network. This means that some might interpret some of the identified weaknesses as software flaws, but other weaknesses can also be seen as configuration issues. That's actually what makes some of our findings hard to 'fix': it's easy to say that someone else is responsible for properly ensuring client isolation :) Hence also hard to really assign CVE(s).
One of the main takeaway issues, in my view, is that it's just hard to correctly deploy client isolation in more complex networks. I think it can be done using modern hardware, but it's very tedious. We didn't test with VLAN separation, but using that can definitely help. Enterprise devices also require a high amount of expertise, meaning we might have missed some specialised settings.. So I'd recommend testing your Wi-Fi network, and then see which settings or routing configurations to change: https://github.com/vanhoefm/airsnitch
CVE are handed out like candy in Java land for artifacts that have code that only opens up a vulnerability when another package is available and the first artifact is misconfigured. So I think you would be fully in your right to claim a CVE and list all affected versions of devices/firmwares there.
I think you could apply specific CVEs to specific devices + setting combination, as:
CVE 1 : router brand X software version Y.Z configured with client isolation does not provide sufficient isolation that it cannot be broken with air snitch.
CVE 2 : router brand A software version B.C configured with client isolation does not provide sufficient isolation that it cannot be broken with air snitch.
etc.
Do separate VLANs behind the different SSIDs provide protection?
I would guess that the VLAN separation should prevent it, but perhaps there are implementation errors on the VLAN implementation inside of individual brands of routers?
Inter-VLAN routing shouldn't be done at the wifi access point, packets would need to be tagged coming out of the wifi AP and switched upstream, unless I'm mistaken about this.
[delayed]
That should definitely help. You still have to double-check the IP routing tables between the VLANs, but most of the time, that should prevent attacks between SSIDs.
Hi! In the case of accessing the private Enterprise SSID, was the network VLAN isolated or some other type of virtualization of the bssid?
Thanks for your work on the topic! This is quite interesting!
When testing our own Enterprise devices, VLANs were not used. This was done to understand the impact of client isolation on its own.
For the university networks that we tested, I'd have to ask my co-author. But perhaps my other comment can further contextualize this: https://news.ycombinator.com/item?id=47172327 Summarized, I'm sure that it is possible to configure devices securely, and VLANs can play an important role in this. But doing so is more tedious and error-prone than one may initially assume, e.g., there is often no single setting to easily do so.
Much of (if not the vast majority of the 'worthwhile') traffic you're intercepting is still encrypted packets though.
Not to minimize the recon value of the plaintext stuff. But not really fair to say you're 'bypassing' any encryption but for the WPA-specific kind.
People who use or rely on client isolation want to prevent inter-client attacks, for whatever reason. We show that this can often be broken. This can be problematic when you have older hardware in your network that is rarely updated, and many then rely on client isolation to mitigate attacks. If everything is encrypted and properly patched, then our attack indeed has less impact, but then there also wouldn't have been a good reason to use client isolation in the first place ;)
[delayed]
That's my read as well. It's bad for places that rely on client isolation, but not really for the general case. I feel like this also overstates the "stealing authentication cookies": most people's cookies will be protected by TLS rather than physical layer protection.
Still an interesting attack though.
I think that places that rely on client isolation might be the general case - every public space that has a guest network - e.g. retail stores, doctor’s offices, hotels, hospitals - is probably using client isolation on their wireless network.
Access points frequently have multiple BSSIDs even if just for broadcasting on 2.4 and 5 at the same time. Any multiple AP scenario will have them regardless. Couple that with weak duplicate MAC checking and shared GTK (WPA2-PSK) and the attack becomes trivial. I imagine old hardware will be broken forever. Especially pre 802.11w.
That’s my read as well. It’s not good, but it’s not nearly as bad as the headline makes it sound.