802.11 and swiss cheese

COMMENTARY--There is no doubt that 802.11b -- the technical name for products also known as AirPort, Orinoco, Aironet, et al--is a life-changing technology. All of a sudden companies don't have to string as many cables through their offices to provide connectivity. Small offices, home offices, and even just plain homes, are all beneficiaries as well, since you can set up an access point somewhere in the house, ideally hidden from plain sight, and still engage in e-mail and wander around the Web.

The problem is that, unlike a piece of cable that you have to get physical access to in order to connect, it's comparatively easy to get near enough to a wireless access point to get good signal strength. Say, in a café across the street.

OK, but just because you're in the radio footprint of an access point doesn't mean you can do anything useful with that wireless network, right? Well, maybe.

Placebo
Even the most user-friendly access points come with basic security facilities. These security features give the appearance of protecting a wireless network in two ways: making the traffic that flies through the ether undecipherable by outsiders and making the access point, well, inaccessible to anyone unauthorized.

The encryption part is accomplished by defining a password that the access point and all its clients share. One known weakness is that the encryption scheme--called WEP--uses a key length of 40 bits, so it's well behind the state of the art. However, I wouldn't be nearly so perturbed if one really needed to brute force attack the full key length. One doesn't.

But wait, there's more. It also turns out that the parts of 802.11's security not related to encryption are also flawed and can be compromised. In short, even if you put all three available security mechanisms--WEP encryption, MAC-based access control, and closed networks--a smart and determined evildoer can still compromise your network.

Lots of homework was done
At least as far back as last October, the IEEE 802.11 committee knew about the security flaws in 802.11 (zipped Word file) and was starting work to fix them. Earlier this year, researchers with the Isaac project at UC Berkeley publicized quite a few problems with WEP. Upon reviewing this work and the design of 802.11's security, respected Bell Labs security researcher Steven Bellovin was quoted in the Wall Street Journal on February 5th as saying that there were some "real howlers" in the design.

WECA, the Wireless Ethernet Compatibility Alliance, promptly issued a formal response (PDF file) after the Berkeley researchers announced their findings. Unfortunately, this response evoked little more than the lightbulb joke whose punchline is "none--they redefine darkness as the standard." The response spent more time focusing on semantic quibbles and how hard it is to perform the attacks than admitting there were fundamental flaws in the protocol in the first place.

Adding to the UC Berkeley findings, a group of researchers at the University of Maryland published a paper of their own (PDF file) outlining even more vulnerabilities in 802.11.

Both the quality and quantity of examination of 802.11's security leaves little doubt about its significant shortcomings.

The sky isn't falling. Yet.
It's worth pointing out these many vulnerabilities that make 802.11's security reminiscent of swiss cheese are manageable now that they're understood. There can no longer be any false sense of security.

It requires a determined attacker to put significant effort into compromising an 802.11 network secured with today's technologies. It's not trivial. That said, all it takes is one person to automate such an attack, make the necessary code publicly available, and suddenly the tools to mount a successful attack will be available to every script kiddie on the planet.

This is the reason why it isn't sufficient for the various 802.11 hardware vendors--not to mention WECA--to handwave and say that it's too much effort to attack such networks in the first place and therefore we should not worry and should be happy. Furthermore, I heard from multiple companies that their belief was that anyone with intellectual property to protect is using a VPN running over 802.11 anyway.

That's just dandy. We're effectively being told that unless we are a large enterprise with a dedicated IT staff and the necessary infrastructure to set up VPN servers and associated folderol we're not worthy of properly designed and implemented security. A flawed system is considered sufficient.

802.1x and AES
While things look grim at the moment, work is already well underway to plug the many holes. One help will be the introduction of 802.1x, support for which will be built into Windows XP when it ships later this year. 802.1x adds improved authentication and access control to Ethernet networks, including 802.11. 802.1x will significantly reduce the vulnerability of WEP to attackers trying to compromise network data.

But even with 802.1x, WEP remains broken. To that end, 802.11e is being worked on, which, among other things, is slated to add 128-bit AES encryption to fix the 802.11's encryption woes.

The current guesstimated timetable for 802.1x support is later this year, and first 802.11e products will likely ship somewhen in 2002. When asked, hardware vendors get really cagey about whether these improvements will be firmware upgrades or whether existing 802.11b hardware will have to be replaced completely. Understandably so since if the fixes require hardware replacement, it might well put quite a damper on 802.11 hardware spending today.

Part of the problem in determining whether existing hardware is upgradable is that current prototype implementations of 802.1x rely on an authentication server in addition to the 802.11 access point. It's far from clear how--or even whether--it will be possible to add 802.1x to stand-alone access points. If not, Home/SoHo users are unlikely to benefit from 802.1x for their 802.11 networks.

Non-corporate users are customers, too
This entire wireless security fracas is as much a war of marketing copy as anything else. The IEEE standards committee did 802.11 users a great disservice by allowing such a flawed security architecture to be approved as a standard. After these flaws became publicized, instead of fessing up, the committee and WECA tried to market their way out of the corner they'd painted themselves into. Neither of these actions reflect well upon the IEEE.

However, the biggest lesson to be learnt should be among users. Just because a new and staggeringly useful gadget offers you capabilities that appear to be security features doesn't mean that they actually provide substantive security. The 802.11 hardware vendors are primarily concerned about large corporate customers; they seem far less interested in how individual customers with a base station and one or two 802.11 cards are served.

If you've already bought 802.11 hardware--especially if you are a non-enterprise user--I encourage you to contact both your hardware manufacturer and your OS vendor and find out what they plan to do about providing adequate security for your wireless network. Until these companies realize that their customers aren't apathetic and do care about the security of small and home networks, they are unlikely to expend much--if any--effort in meeting those needs.

Stephan Somogyi feels more than a little vindicated after hearing that IEEE1394 will be supported in Windows XP and USB 2.0 won't. He also wishes that Apple and Agere would agree on who's responsible for adding Orinoco card support to OS X and get on with implementing it.