Researchers Uncover North Korean Operatives in Fake Crypto Startup

In a recent investigation, security researchers created a fictitious cryptocurrency startup to expose North Korean operatives infiltrating the tech industry. By advertising developer positions, they successfully hired three individuals suspected of being North Korean IT workers. Each was provided with virtual machines (VMs) that recorded all activities.

During the onboarding process, discrepancies emerged. The first hire claimed residency in Pasadena, Texas, but submitted a California driver’s license and a New York bank account. Metadata analysis revealed the image had been processed using Google Gemini, with a SynthID watermark indicating AI-generated content. The second applicant provided a Texas license, a valid Social Security number, and a Kansas City bank account. The third submitted a New York license belonging to another individual and a genuine iPhone 15 photograph with GPS data removed.

These operatives gained legitimate employee accounts, granting them access to source code and internal systems. A joint alert issued on July 31 highlighted that North Korean IT workers seek contracts to funnel earnings back to North Korean agencies. Employers are advised to be vigilant for documents altered using image editing software.

In April, the U.S. Department of Justice sentenced two individuals involved in placing North Korean workers in over 100 U.S. companies using stolen identities, generating more than $5 million for North Korea. While Google’s Gemini app can detect SynthID watermarks, it only identifies content created or edited by Google’s AI models, leaving room for undetected alterations by other tools.

This operation follows a previous endeavor by threat intelligence firm BCA LTD, the NorthScan research initiative, and sandbox provider ANY.RUN. In late 2025, they posed as facilitators renting out identities. This time, they established a fake decentralized finance (DeFi) protocol named Ballena Azul. A recruiter sourced the first developer via GitHub, who then referred two others.

Upon hiring, each operative underwent standard procedures, including interviews and contract signings, before receiving access to work VMs. Initial activities involved system reconnaissance, such as running diagnostic tools and verifying connection origins. One operative installed Chrome Remote Desktop, syncing a personal Google account to the sandbox, thereby exposing browsing history, saved passwords, and installed extensions. GitHub access was also noted.

The tools observed differed from previous operations. The researchers identified the use of 2fa.cn for two-factor code transmission, whereas earlier campaigns utilized authenticator.cc and otp.ee. Outlook.com was also employed, replacing prior reliance on Gmail. Additionally, browser extensions like AIApply, Final Round AI, Simplify Copilot, and ChatGPT prompt tools were detected. Infrastructure was traced to Vultr and Gorilla Servers, with AstrillVPN exit nodes in use.

To mitigate such threats, researchers recommend periodic identity verification beyond initial hiring, in-person checks for remote employees, recruiter training, and blocking AstrillVPN. The July 31 advisory further emphasizes the need for vigilance against these sophisticated infiltration tactics.

This incident underscores the evolving strategies employed by state-sponsored actors to infiltrate organizations. By leveraging advanced social engineering and AI tools, these operatives can gain legitimate access to sensitive systems. Organizations must enhance their vetting processes and remain alert to such sophisticated threats to safeguard their assets and data.