Although Microsoft patched the ANI bug in Windows a week ago, that's not stopped attackers from boosting the number of Web sites serving exploits or tweaking exploits to make them more efficient.

"Based on past history, they'll have a pretty high success rate for quite a while," said Dan Hubbard, Websense's head of research, of attackers using ANI exploits. "That success rate goes down with time, but a patch is not the end of the story."

Past vulnerabilities that have been the target of major exploit campaigns -- notably the Windows Metafile (WMF) bug of late 2005 and early 2006 -- still account for 15 percent or more of active attacks, says Hubbard. "And that's more than a year after [the vulnerability was patched]."

As of Tuesday, according to Hubbard, over 2,000 Web sites either purposefully malicious or compromised by criminals are hosting exploits against the ANI file bug in Windows 2000, XP, Server 2003, and Vista. "That's bigger than WMF," he said. "The number of sites serving ANI exploits is larger than the number a week or so after WMF started."

Most of those sites are related to one of two major attacker groups. The first, out of China, is widely believed to be the first to use exploits, which were detected about a week before Microsoft released the emergency patch. These attackers, according to Hubbard, were using compromised Web sites to hijack users' log-on credentials for online games such as "Lineage," popular in Asia. Most of the servers compromised by this group are in China.

A second group from Eastern Europe is behind the bulk of the rest. This gang added ANI exploits to those it already used -- including WMF and VML (Vector Markup Language) exploits -- to grab control of servers primarily in the U.S. The second group's goal, said Hubbard, was more straightforward crime: "They mainly deal with information-stealing Trojans," he said.

As Websense tallies up the compromised sites, other security vendors have been finding new exploit-building toolkits and the stealthier malware some of those tools create.

In a Trend Micro Inc. blog, researcher Jonell Baltazar posted screen shots of toolkits for creating ANI exploits. "These toolkits makes it easier for a script kiddie to create a malware of his own," said Baltazar.

Over at McAfee's Avert Labs, meanwhile, researcher Geok Meng Ong spelled out obfuscation techniques that some ANI exploits were using to sneak by defenses. In one sample, the code of a toolkit-generated exploit used random tags to avoid detection as an .ani (animated icon) file. Most image viewers including Internet Explorer parse them without any problems," said Geok.

"We [also] found common ANI headers that were modified and redundant noise [extra white space in the code], in an attempt to circumvent detection in most traditional content filtering and anti-virus products," said Geok.

In other words, ANI, though patched, is just getting started.

"So far we haven't seen any activity of ANI exploits grabbing systems for botnets, but we expect to see that next," Hubbard said.