Tilde-D Detection Focuses On Coding Anomalies

Last month, researchers at Kaspersky Labs found stronger evidence that the same programming platform was used to create Duqu, an espionage Trojan, and Stuxnet, an attack launched in 2009 on Iran's nuclear processing capability.

The two attacks seem quite different at first glance. Stuxnet aimed to cripple the industrial systems in a specific target that contained a highly unique environment, while Duqu is designed to steal information from a variety of systems. Stuxnet spread widely, driven by the need to spread into a highly secure network, while only a dozen instances of Duqu have been found.

Yet a number of key components within the two threats are highly related -- so closely that it's very likely they were created by the same highly automated tool, a creation platform that researchers have dubbed "Tilde-D" for its use of "~d" at the beginning of file names. The similarities have been used by researchers at the Laboratory of Cryptography and System Security (CrySyS), a group within the Budapest University of Technology and Economics, to create an open-source tool for detecting Duqu and other code created by the Tilde-D kit.

"Generally speaking, companies can run our tool ... on all computers and then filter the resulting information to pinpoint infections or suspicious computers," says Dr. Boldizar Bencsath, lead researcher on the Duqu project at CrySyS.

The tool, currently named the Duqu Detector Toolkit but which can detect other Tilde-D creations, has four components that look for specific aspects of the malicious programs. The first component looks for signs of the loader driver, a kernel component that injects malicious code into the system during exploitation. A second component looks for unexplained entropy, a sign that files are encrypted, among INF files. In real-world tests, encrypted Duqu files showed an entropy of 0.9, much higher than the tool's threshold of 0.6. A third component looks for undeleted temporary files left by Duqu, while the fourth component inspects PNF files that have no corresponding INF file.

While it would seem simple to get around some of the detection methods, the creators of the Tilde-D programming platform will have to change the way their tool creates code, says Bencsath. Each change requires a great deal of testing and then must be pitted against antivirus software and the Duqu Detector Toolkit.

"Basically, they can avoid detection by modifying the code, but to achieve that goal they need to change the way they use the tool," he says.

Yet the Duqu Detector Toolkit has its own issues as well. Tools that attempt to detect malicious software by looking for specific anomalies on a system can be prone to false positives, says Roel Schouwenberg, senior researcher with Kaspersky Labs.

"There are always methods to avoid detection, though the issue in this case is not detection -- it's detection without false positives," he says.

Schouwenberg also pointed out that all signs point to a state actor with a budget of tens of millions of dollars as the creator of Stuxnet. Even if revamping the Tilde-D programming platform to avoid detection requires a great deal of resources, the group has already shown it has the capability to do that.

With the tool in the open-source world, the attackers have another advantage, as well, says Richard Wang, manager of SophosLabs U.S.

"At the end of the day, the bad guys have a big advantage: They can rewrite a piece of malware until they find a variant that does not trigger any defense," he says.

As part of a defense-in-depth approach to security, companies should make sure they have the basics, such as antivirus, good patching processes, and network defenses. Next, adding log management and correlation technologies can help detect persistent attacks that will not be caught by other means. Finally, hiring intelligence services that look for command-and-control servers siphoning off information from clients' servers can sometimes catch what a company has missed.

Like the Duqu Detection Toolkit, many of these tools highlight potential threats for security teams to investigate, says Wang.

"A lot of these technologies are about flagging," he says. "They are not going to pop up and say, 'This is an advanced threat on your network.'"

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