A new injection attack that redirects users' Web search queries is in the wild, and researchers at Websense believe it may have already affected more than 40,000 sites.

In a blog posted yesterday, Websense researchers indicated that more than 40,000 legitimate sites have been compromised with "obfuscated code that leads to a multilevel redirection attack, ending in a series of drive-by exploits which, if successful, install a Trojan downloader on the user's machine."

When users visit one of the infected sites, they are redirected through a series of different sites owned by the attacker and brought to the final landing page containing the exploit code, the researchers say. The final landing page records the visitor's IP address.

When the site is visited for the first time, the user is directed to the exploit payload site. But if the user returns from the same IP address, he is simply directed to the benign site of Ask.com, the researchers report. This one-time download strategy may make the redirects less obvious and harder to detect, they say.

According to a spokesman, the labs first detected what appeared to be benign redirects embedded in compromised Web sites that sent users to Ask.com. "At that time, it seemed likely that hackers were looking to compromise as many sites as possible, getting their foot in the door before activating the campaign with a redirect to a malicious payload site," he says. The attackers used polymorphic code to avoid detection in these early stages.

Now the researchers understand that the malicious campaign actually began simultaneously with the Ask redirect, and the malicious payload site ninetoraq has been infecting users with malware.

Once the user's computer has been redirected from a compromised site to ninetoraq, the site attempts multiple exploits through obfuscated code targeting vulnerabilities in MDAC, AOL SuperBuddy, Acrobat Reader, and QuickTime, the spokesman says. If it finds an open hole, it drops a malicious PDF file or a Trojan that is designed to steal the user's information.

Most antivirus applications will not detect either one of these pieces of malicious code, Websense says. One of the exploits is detected by only three of the 41 most commonly used AV programs.

"The obfuscation code injected into these legitimate Web sites is somewhat random, but the deobfuscation algorithm is consistent amongst all the infections," the researchers say. "The algorithm uses the JavaScript method 'String.fromCharCode' to convert a chunk of decimal values to a string. The string obtained after deobfuscation is an iFrame that eventually leads to an exploit site."

The Websense researchers say the new attack is distinct from Gumblar or Beladen, two other injection attacks that have been redirecting users' search queries in the past month. It is possible that the same hackers might be developing the different attacks, they say.

Have a comment on this story? Please click "Discuss" below. If you'd like to contact Dark Reading's editors directly, send us a message.