The flaw was made public Thursday, when security researcher Nikolaos Rangos posted details of the vulnerability to the Full Disclosure security mailing list. By sending a specially crafted HTTP request to the server he was able to view and upload files on the machine. The attack takes advantage of a bug in the way that Microsoft's software processes Unicode tokens, he said.

The vulnerability is being used in online attacks, the U.S. Computer Emergency Response Team said Monday.

In a statement, Microsoft said it hadn't heard of any such attacks, but that it was investigating Rangos' claims. "We are working on a security advisory to provide customers with guidance," the company said Monday.

The bug affects IIS 6 users who have enabled the WebDAV (Web-based Distributed Authoring and Versioning) protocols, used to share documents via the Web.

It gives attackers a way to view protected files on the server without authorization and it could be used to upload files as well, according to Thierry Zoller, an independent security researcher who confirmed Rangos' findings. However, Zoller said he had found no way to use this flaw to run unauthorized software on an IIS server.

Zoller said that IIS 5 and IIS 7 do not appear to be vulnerable to the attack, but that it could affect other Microsoft products that use the WebDAV technology. "Better safe than sorry," he said via instant message, "Disable WebDAV temporarily and wait for Microsoft to patch."

In an e-mail interview, Rangos said that even with WebDAV enabled, Exchange Server running on IIS 6 and SharePoint Server were not affected by the flaw.

Cisco sounded a similar warning. "Administrators of sites that are hosting sensitive information on IIS servers that use WebDAV are advised to put effective mitigations into place immediately because exploit code is publicly available," the company said in a security alert posted to its Web site.