This time two weeks ago, the security community was panicking over the potential damage caused by the WMF vulnerability.
You can argue, as I did, that the mitigating factors were strong and it wasnt as serious an issue as some argued, but clearly the watershed event in getting past it was Microsofts prompt release of a patch last week.
“Prompt release of a patch”—theres a phrase you dont hear referring to Microsoft very often. How was it possible for Microsoft to release a patch in about a week? This is unusual for the company.
How unusual? Consider the data that eEye, a security tool and research company, puts out about vulnerabilities it has reported to Microsoft.
eEye has, as these things go, a long history of reporting severe vulnerabilities and getting credit when Microsoft finally discloses and fixes them. Youd think that Microsoft would take vulnerability reports from eEye seriously as a matter of course.
But these are the dates of the current list of vulnerabilities that have been reported to Microsoft:
- Oct. 17, 2005—Severity: High (Remote Code Execution)
- Oct. 11, 2005—Severity: Medium (Denial of Service)
- Aug. 1, 2005—Severity: High (Remote Code Execution)
- June 27, 2005—Severity: High (Remote Code Execution)
- May 5, 2005—Severity: High (Remote Code Execution)
Criminy! May 5? This is not exactly a prompt response. And yet its not unheard of, far from it.
Microsofts explanation for this staggering lead time has generally been about the necessity of testing patches thoroughly and the companys need to release simultaneously in 20-something languages.
Fair points all, as many other vendors and open-source efforts seem to view testing as something their users should be doing. But eight months? Before too long youll be able to test these patches with carbon-14 dating.
And the response to the WMF unmasks the insufficiency of the whole process. Of course, it didnt need this episode to be unmasked; Microsofts slow response was well-known in the past. But never has it responded so quickly to a zero-day attack.
If there is a certain amount of minimal overhead time built into the development and testing process for patches, clearly its not a large amount, and no larger than about a week. And not just any week: The WMF patch testing happened over a holiday weekend!
The Partly Line
So we can say, independent of any issues related to specific patches, that it can be done quickly. What about the specifics?
I could offer some speculation about this particular issue and why it was a relatively simple one that could be disposed of quickly: The fix was obvious, in that it simply disabled support for a feature that turned out to be inherently dangerous.
Once the decision was made to disable a feature the rest was easy.
Also, Microsoft had the advantage of WMF being a largely obsolete format, and this being a relatively obscure feature of it. Even if Microsoft assumed that it had to completely break the feature, it could just declare that the implications were worth living with.
Normally, one would think, this would be an ominous decision that would ripple up and down through management at the company, but if that happened in this case, it too happened quickly. (Another interesting question is how this flaw could have escaped notice in previous security audits of the Windows code. Someone should be made to explain themselves.)
Also, the position Microsoft takes in its security bulletin that the vulnerability is critical for Windows XP, 2003 and 2000, but not for Windows 9x, is inconsistent and probably disingenuous.
The bulletin states, “Although Windows 98, Windows 98 Second Edition, and Windows Millennium Edition do contain the affected component, the vulnerability is not critical because an exploitable attack vector has not been identified that would yield a Critical severity rating for these versions.”
In all fairness, the same might be said about Windows 2000. It too has no Picture and Fax Viewer, the default component that handled WMF files in Windows XP and Server 2003. But all versions of Windows are exploitable through vulnerable applications; I was able to exploit on Windows 2000 with Microsoft Word (the Insert Picture command) and Im sure the behavior would be identical on Windows 98. Why the exception for Windows 2000?
Probably because its a mainstream product with a massive degree of corporate adoption, and Microsoft is trying to get all customers to move away from the earlier versions. So the vulnerability in Windows 98 may be more critical for customers than for Microsoft.
And this is the real secret to the situation: Microsoft listens to its large customers and takes their concerns seriously. Im sure that the Fortune [insert small number] was on the phone to Microsoft demanding prompt action on the WMF issue, and whaddayaknow, prompt action happened.
And large customer demands are the only thing that will normalize a more acceptable window of action by Microsoft for credible vulnerability reports. The days of claiming that eight months is a reasonable test and development period are over.
Security Center Editor Larry Seltzer has worked in and written about the computer industry since 1983. He can be reached at larryseltzer@ziffdavis.com.