Wednesday, October 29, 2008

New Windows Bug (MS08-067) Differs from 2006 Flaw

Via NYTimes -

Contrary to speculation, the bug Microsoft patched unexpectedly last week is not closely related to one the company fixed more than two years ago, a company security expert said Monday.

The vulnerability in the Windows Server service that Microsoft patched with an "out-of-cycle" fix last Thursday is unrelated to another hole in the same service that the company plugged with a patch in August 2006, said Michael Howard , a principal security program manager with the company.

Howard, perhaps best known for co-authoring the book Writing Secure Code , works in the Security Development Lifecycle (SDL) group at Microsoft; SDL is also the company's name for the process it uses to generate more secure software.

"I spent a good chunk of the morning going over the analysis [of the bugs]," said Howard at mid-day Monday. "The two bugs are actually quite different. Although the effect is the same, the way you pull off an exploit is very different."

The two vulnerabilities are so different, he added, that it was no surprise that company developers didn't find the most recent one when they looked at the Windows Server service code two years ago. "[This one] is a really hard bug to spot," Howard said.

Microsoft patched the Windows Server service in August 2006's MS06-040 . Like the one posted Thursday , that bulletin calls out the remote procedure call (RPC) code in the service as the location of the bug.

To back up his claim that the newest flaw is easy to overlook, Howard cited a Twitter message three days ago from Alexander Sotirov, a noted independent security researcher. "I had the vulnerable function decompiled and fully commented back in 2006 when I was reversing MS06-040, but I just didn't see the bug," said Sotirov , who was one of a pair of researchers who helped Shane Macaulayhack Windows Vista SP1 during a March contest.

Howard acknowledged that parts of Microsoft's SDL process, specifically the fuzz tests thrown at the affected code, had failed. But he was generally upbeat about how well SDL did its job.

"At the end of the day, I'm not unhappy with how SDL performed," he said. "We have two goals. One is to reduce the number of vulnerabilities, and the second is to reduce the severity of those we miss. You're never going to get everything. Windows Vista and [Windows] Server 2008 [users] were protected; SDL reduced the seriousness of the vulnerability for them.

"So I think SDL did succeed here," Howard said.

No comments:

Post a Comment