Microsoft confirms typo responsible for IE exploit
Rogue '&' character blamed.
By Gregg Keizer, Computerworld (US) | Computerworld US | Published: 11:12, 30 July 2009
Redmond has confirmed that a single superfluous character in its own development code is responsible for the bug that has let hackers exploit Internet Explorer (IE) since early July.
A pair of German researchers who analysed a vulnerability in a Microsoft-made ActiveX control came to the same conclusion three weeks ago.
"The bug is simply a typo," Michael Howard, a principal security program manager in Microsoft's security engineering and communications group, said in a post to the Security Development Lifecycle (SDL) blog. Howard, who is probably best known for co-authoring Writing Secure Code, went on to say that the typo -- an errant "&" character - is the "core issue" in the MSVidCtl ActiveX control.
That video-streaming control was created by Microsoft using a modified version of an older edition of a code "library," dubbed Active Template Library (ATL), that Microsoft admitted contained multiple vulnerabilities. Microsoft also patched Visual Studio, the company's development platform that contains ATL. Those patches, however, do not automatically fix software that was developed using the buggy ATL. Instead, vendors - Microsoft as well as third-party firms - must use the patched Visual Studio to recompile their code, then distribute the new, secure software to users.
Howard said that the bug in the MSVidCtl ActiveX control was introduced by an internal version of ATL, not one that was available to outside developers.
Others at Microsoft drew a direct line between the ATL bug Howard described to the public exploits hackers have been using for much of this month, including drive-by attacks conducted from thousands of compromised websites.
"This public exploit took advantage of the fact that MSVidCtl uses a modified version of vulnerable ATL headers," said Fermin Serna, an engineer in the Microsoft Security Research Center (MSRC), in a blog post. "In this specific instance, the vulnerability allows an attacker to corrupt memory which may lead to a remote code execution," added Serna.
