APT29 Cyberspies Use Domain Fronting to Evade Detection

27.3.2017 securityweek APT
The Russia-linked cyber espionage group known as APT29 has been using a technique called “domain fronting” in an effort to make it more difficult for targeted organizations to identify malicious traffic, FireEye reported on Monday.

Domain fronting is a censorship bypassing technique that involves disguising traffic to make it look as if it’s going to a host allowed by the censor, such as Google, Amazon or CloudFlare. Open Whisper Systems recently implemented the technique to help Signal users in Egypt and the United Arab Emirates bypass government censorship.

According to FireEye, the technique has been used for at least two years by the threat actor APT29, which is also known as The Dukes, Cozy Bear and Cozy Duke. The group is believed to be behind the recent election-related attacks in the U.S. and a campaign targeting high-profile organizations in Norway.

APT29 has used the Tor anonymity network to communicate with infected machines, which could be considered suspicious by some defenders. In order to disguise Tor traffic as apparently legitimate traffic, the cyberspies used Meek, a Tor plugin that implements domain fronting and allows users to send traffic to Tor inside a harmless-looking HTTPS POST request to google.com.

In its attacks, APT29 used a PowerShell script and a .bat file to install the Tor client and the Meek plugin on the targeted system. They leveraged an exploit involving the Sticky Keys accessibility feature, where they replaced the legitimate executable with the Windows Command Prompt (cmd.exe) file. This provides the attacker a shell that they can use to execute commands with SYSTEM-level privileges, including to add or modify accounts.

The script that executes the Sticky Keys exploit also creates a Windows service named “Google Update” to ensure that the backdoor remains even after the system has been rebooted.

“APT29 adopted domain fronting long before these techniques were widely known,” said FireEye’s Matthew Dunwoody. “By employing a publicly available implementation, they were able to hide their network traffic, with minimal research or development, and with tools that are difficult to attribute. Detecting this activity on the network requires visibility into TLS connections and effective network signatures.”