The Month of Kernel Bugs project released details for taking over a computer with an affected Broadcom Wi-Fi driver: eWeek reports that Jon Ellch ("Johnny Cache") discovered the flaw earlier this year. Ellch provided details to Broadcom, which released updated drivers to its manufacturing partners, only some of which have apparently distributed end-user fixes to their customers in the form of driver updates or as part of larger software updates. Yesterday's release includes a Metasploit module for automated, relatively exploitation of this flaw which leads to complete control of a computer.

The exploit results from Broadcom's driver improperly parsing a response to a request for the network name (SSID or service set identifier) from a nearby device. These probe requests are sent out regularly by Wi-Fi adapters that are not associated with a network; they can also be sent by associated adapters updating the computer's picture of what networks are active in the vicinity. A malformed probe response--in the Metasploit module, this appears to be a crafted 93-byte name--can overflow a buffer and allow "arbitrary kernel-mode code execution," which can be turned into root-level ownership of the operating system. As long as the Wi-Fi radio is turned on, any computer with the right unpatched firmware is vulnerable. An attack requires a malicious user within antenna range of a vulnerable machine; the exploit cannot be executed remotely.

Any Wi-Fi adapter with Broadcom driver BCMWL5.SYS version 3.50.21.10 under Windows XP, and potentially under Linux and FreeBSD distributions that can make use of Windows Wi-Fi drivers. The SANS Institute reports that this driver version is known to be vulnerable, but that others may be as well. The driver in question is bundled with computers made by a variety of vendors, including HP, Dell, eMachines, Gateway, and others, as well as third-party wireless cards from Linksys and others. A complete list of affected computers and chipsets (if limited to certain chipsets) is not available at this writing. According to what's known today, only Linksys's patches for its WPC300N Wireless-N wireless card contains the updated Broadcom firmware to fix this vulnerability.

The so-called Zeroday Emergency Response Team (ZERT) said there's no simple way for them to apply their usual ZERT methodology of releasing temporary patches for zero-day exploits in this case. Zero-day exploits are flaws that are announced at the same time as code is generally available to take advantage of the flaw. ZERT tries to release code to defeat these exploits to prevent short-term problems, but recommends vendor-supplied patches as they become available.

You'll recognize Ellch's name from the August through October soap opera in which Ellch and colleague David Maynor first apparently stated that they had found weaknesses in certain drivers that allowed non-associated Wi-Fi adapter attacks through weak drivres, and either implied or stated that this included Apple's Mac OS X operating system and associated Wi-Fi drivers.

The two researchers later couldn't be pinned down on precisely what they did say, and Apple denied that the researchers provided them with information that led to patches later released for OS X to fix what Apple described as flaws that hadn't yet been exploited. (Those flaws involved malformed frames, a higher-level and more generic problem; they don't appear to be identical in nature to this Broadcom vulnerability. Ellch's methodology in discovering the flaw was apparently the same.)

Ellch and Maynor have never publicly released the information they say they provided to Apple, and Maynor's employer SecureWorks put a kabosh on discussion by restricting him from speaking about it publicly (he dropped out of a scheduled Toorcon presentation with Ellch), while agreeing to work with Apple and CERT on vague coordination issues. (Ellch doesn't work for SecureWorks, but appears bound by the same confidentiality agreements that Maynor has observed.)

Some commentators and security experts said that Ellch and Maynor had the goods; others thought they were posers. I tried to stay consistent on this topic, which is difficult. I don't believe Apple was lying, based on both public and private information I obtained; and I have had a hard time believing, too, that Maynor and Ellch were simply seeking publicity through fabrication. That seemed unlikely given their track record.

This cognitive dissonance has led some commentators and columnists to support Maynor and Ellch unconditionally, even without seeing the actual exploits or any of the code, while some Macintosh focused writers have stated unequivocally that they think the whole matter was hand waving. (John Gruber of Daring Fireball and Jim Thompson agreed to buy and give two fresh MacBooks to either researcher if they could show up and crack the machine per their stated exploit.)

In this current case, it looks like Ellch gets to smell like a rose with no dispute over process, proof, or results. First, he revealed the exploit privately to Broadcom with sufficient advance word that Broadcom could create a patch that Linksys was able to incorporate by Nov. 6. Second, his disclosure is fully documented. Third, the disclosure has an excellent social purpose, as well, as it will force manufacturers that may have dallied on providing fixes or full disclosures of this risk (if they knew about it), to push patches out right away. Update: Perhaps not. Broadcom says that Ellch gave them no advance word; Ellch hasn't responded to my query on this.

While this is a serious exploit, it has to be carried out by individuals, even individuals with high-gain antennas. Because the vector doesn't work over the Internet or over local networks, only within the range of active Wi-Fi adapters accepting probe responses within reach of a malicious user, this reduces the scope of the number of possible machines infected. A crazy black-hat wardriver might be able to drive around a city and own machine after machine, true.

What can you do in the meantime with an affected driver? Unfortunately, the only option for true security is finding an Ethernet cable and disabling your radio.

Update on Monday: Still no details on how an unassociated Wi-Fi adapter can be hijacked through the exploit; perhaps the payload of the exploit enables peer-to-peer networking thus allowing a remote user to then access the cracked machine?

Broadcom released a statement this morning that they started auditing their code after what they describe as a "widely publicized vulnerability in certain wireless implementations that was announced at the annual Black Hat conference," which would be the alleged Apple Wi-Fi driver flaws. The company said they found some vulnerabilities via that audit and developed new testing tools that will help them avoid similar holes in the future.