Hackers Target US AI Experts With Customized RAT

A possible Chinese threat actor is using a variant of the Gh0st RAT malware to steal information from artificial intelligence experts in U.S. companies, federal agencies and academia.

See Also: Global Threat Report 2024: Executive Summary

Named UNK_SweetSpecter, the threat actor was interested in a “specific software,” targeting experts using an AI-themed email phishing lure to spread the remote access Trojan, Proofpoint researchers said in a Thursday blog post.

On the criminal group’s target list were fewer than 10 technical personnel, all of whom had connections to a “leading U.S.-based AI organization,” the researchers said, without identifying the company. It is likely that the bad actor was looking to gain access to nonpublic information about generative AI, they said.

The researchers first spotted the campaign earlier this month but said there isn’t “enough telemetry to link the malicious activity to any known nation-state threat actor or objective.”

SugarGh0st RAT was first reported by Cisco Talos in November, when it identified a Chinese-speaking threat actor carrying out a cyberespionage and surveillance campaign against Uzbekistan and South Korean government officials.

The customized variant of Gh0stRAT surfaced in 2008 when a Chinese hacker group called C. Rufus Security published its source code in public, which led to nation-state and criminal actors using it in several threat campaigns. Hackers have used SugarGh0st RAT to target a U.S. telecommunications company, an international media organization and a South Asian government organization via publicly available email addresses in most cases.

SugarGh0st carries several of Gh0st RAT’s capabilities, including the ability to enable full remote control of the infected machine, real-time and offline keylogging, spying via the system webcam and downloading additional malware.

It’s better than the original in other ways. For one, the new variant is designed to have reconnaissance capabilities for particular objectives, which enables it to search for specific Open Database Connectivity registry keys that hackers could use to move laterally and exfiltrate data. The updated version can also load and execute malicious code from library files with specific file extensions and function names. Remote operators can use the new variant to issue custom commands via the command-and-control interface.

In the latest campaign, Proofpoint said, the hacker used a free email account to send the targets an attached zip archive, which they claimed contained questions about problems with an AI tool. They were seeking assistance from the relevant technical personnel. When downloaded, the attached zip file dropped in a compromised shortcut file similar to what Cisco Talos found in its original SugarGh0st analysis last year. The shortcut file deployed a JavaScript dropper with a decoy document, an ActiveX tool for sideloading and an encrypted binary, all encoded in base64. The SugarGh0st malware then completed the infection chain by communicating with an attacker-controlled C2 server.

The campaign report comes at a time when the U.S. government is restricting Chinese access to gen AI software and tools. The Department of Justice recently indicted a Google software engineer for stealing AI secrets to deploy at two AI-related technology companies in China, including one where he was the founder.

“It is possible that if Chinese entities are restricted from accessing technologies underpinning AI development, then Chinese-aligned cyber actors may target those with access to that information to further Chinese development goals,” Proofpoint said.