Skip to content
Chimera readability score 78 out of 100, Expert reading level.

Suspected North Korean threat actors are targeting developers with fake job offers and “coding assignments” that lead to the deployment of cross-platform malware for cryptocurrency and credential theft, Proofpoint reported Monday.The threat cluster, tracked as UNK_DeadDrop, shows similarities to Contagious Interview and sent more than 250 emails across nearly 100 target companies over a six-week period between April and May 2026. The emails mostly targeted technology, education, business services, financial services, entertainment/media companies and telecommunications companies in the United States, with a particular focus on the cryptocurrency industry.In the emails, the attackers pose as recruiters from legitimate companies such as the decentralized finance company Ondo Finance, the telehealth company Nourish and the Web3 and AI talent agency Hypen Connect, offering the target to apply for a software development role and complete a coding assignment.In some cases, the attackers also presented as fake cryptocurrency and AI startups, with names like Pulsynk and Trixauvex, requesting the target for a peer-review of their code. In either case, the attacker included a link to a GitHub or GitLab repository with instructions to clone the repo to Visual Studio Code (VS Code) or Cursor code editors. The repositories include a hidden tasks.json file that abuses VS Code and Cursor’s task automation abilities to automatically execute malicious files when a certain project folder is open in the editor. VS Code will prompt the user to approve the task execution while Cursor does not display a prompt, Proofpoint noted.
Related reading:
The malware acts differently depending on whether the victim is using Windows, Linux or macOS.On Linux and macOS systems, the attacker leverages an open-source command-and-control (C2) framework called Overlord, deploying Go binaries with remote access trojan (RAT) capabilities that establish a persistent WebSocket connection to the attacker’s servers.On Windows systems, the attack chain runs as JavaScript within the code editor’s Electron process and performs a single infostealer operation without persistence, Proofpoint said.The Overlord RAT first extracts browser wallet extensions and standalone wallet directories and transmits them as a ZIP archive to the C2 server. Five minutes later, it displays a fake system dialog prompt for the user to enter their system password, leveraging a Mach-O binary on macOS systems and the Zenity tool on Linux systems. If the user enters their password, the malware leverages the password to extract browser credentials from Keychain and GNOME Keyring and subsequently relaunches itself as root to perform further Keychain and GNOME Keyring dumps.On Windows, the malicious task launches a VBScript that calls a CMD file, which then decodes an additional embedded script that stages three encrypted payloads. These payloads are decrypted at runtime and serve to facilitate infostealing capabilities.The malware scans for 35 Chromium browser wallet extensions and 18 standalone wallet applications and also uses a Python-based stealer to extract passwords from Chromium and Firefox browsers, and cookies from Chrome, Edge and Brave browsers. The stolen data is sent to the C2 server via an HTTP POST request.The malicious task also installs a malicious VSIX extension that enables persistence on Linux and macOS machines, causing the malware processes to be relaunched every time VS Code or Cursor are opened. This extension is installed on Windows machines but does not re-execute the malicious operation on these machines.Proofpoint found that UNK_DeadDrop mainly emailed victims from domains registered via Namecheap set up with MailHostBox mailservers. Some of the domains were hosted on Vercel and included websites promoting the fake startups that were likely AI-generated, the researchers said.UNK_DeadDrop’s victim targeting, social-engineering techniques and theft of cryptocurrency wallets and credentials overlap with the North Korean Contagious Interview campaign. However, this cluster shows some distinct techniques such as the use of email rather than social media channels like LinkedIn to contact victims, the abuse of task.json auto-execution rather than npm installation, and the use of Overlord Go binaries rather than other malware known to be used by Contagious Interview, such as OtterCookie and FlexibleFerret.“The consistent creation of new GitHub repositories as well as a new malware framework with iterative builds and a stealthy new execution and persistence technique through VSIX extensions demonstrates dedicated resourcing and active development of tooling,” the researchers concluded.
Get daily email updates
SC Media's daily must-read of the most current and pressing daily news
You can skip this ad in 5 seconds

Facts Only

* Threat actors operate under a cluster tracked as UNKDeadDrop.
* The threat actors pose as recruiters from entities such as Ondo Finance, Nourish, Hypen Connect, Pulsynk, and Trixauvex.
* Attacks involve targeting developers with fake job offers and coding assignments.
* Attackers provide links to GitHub or GitLab repositories for malware deployment.
* Malware utilizes a hidden tasks.json file to abuse VS Code and Cursor task automation abilities for malicious execution.
* On Linux and macOS, the attack uses the Overlord Go binaries with remote access trojan capabilities to establish persistent WebSocket connections.
* On Windows, the attack involves JavaScript within the Electron process performing infostealer operations without persistence.
* The malware extracts browser wallet extensions, standalone wallet directories, and system credentials from Keychain (macOS) or GNOME Keyring (Linux).
* Malware stages encrypted payloads in Windows using VBScript calls to CMD files.
* The campaign primarily targets technology, financial services, and telecommunications companies in the United States.

Executive Summary

Threat actors utilize a sophisticated strategy to target developers by masquerading as recruiters from legitimate technology companies and fake AI/crypto startups. This campaign involves sending emails containing fake job offers and coding assignments, which serve as the initial vector for infection. Attackers leverage open-source code repositories (GitHub/GitLab) to deliver malware disguised as development tasks. The resulting attack chain is highly adaptive, employing different methods based on the operating system (Windows, Linux, macOS). On Linux and macOS systems, the attackers use a Go-based Command-and-Control (C2) framework called Overlord for persistence via remote access trojans. On Windows systems, the malware operates through JavaScript within the code editor's Electron process to perform infostealing operations. The ultimate goal of this coordinated effort is the theft of cryptocurrency wallet credentials and system secrets from users.

Full Take

The described attack structure demonstrates a sophisticated exploitation of trust inherent in the developer ecosystem. The reliance on impersonating established, legitimate entities—such as DeFi firms or Web3 agencies—is not merely social engineering; it exploits the professional context where developers naturally accept tasks and deadlines from familiar sources. This setup transforms a legitimate workflow (hiring/coding) into a malicious pipeline for credential harvesting and malware deployment. The technique of abusing internal automation features like `tasks.json` highlights an understanding of how modern development environments operate, shifting the attack surface from traditional file-based malware to functional software mechanisms.
The pivot to utilizing distinct C2 frameworks based on the operating system (Overlord RAT for Unix-like systems versus JavaScript/Electron processes for Windows) demonstrates a dedicated commitment to evading signature-based detection and leveraging native execution privileges specific to each platform. This pattern suggests that the actors are not merely using off-the-shelf malware but actively developing iterative toolsets, evidenced by the consistent creation of new GitHub repositories and evolving frameworks. The operational separation between the initial contact (email) and the final execution (malicious task automation) illustrates a systemic evasion strategy that moves beyond simple payload delivery toward sustained, adaptive persistence within high-value targets.
Patterns detected: ARC-0043 Motte-and-Bailey, ARC-0024 Ambiguity, ARC-0051 Systemic

Sentinel — Human

Confidence

LIKELY_HUMAN (confidence: 0.15)