Attackers are actively exploiting a critical vulnerability in SimpleHelp — CVE-2026-48558 — to deploy a previously undocumented stealer called Djinn. The malware targets Windows, macOS, and Linux, with a particular focus on developer machines and AI coding tools.
SimpleHelp is remote monitoring and management software used by MSPs, IT departments, and helpdesks. Earlier this month, Horizon3.ai published details about the flaw, which allows unauthenticated attackers to create highly privileged technician accounts. At disclosure, roughly 1,000 SimpleHelp servers exposed online were running a vulnerable configuration.
The attack chain works like this: the threat actor exploits the authentication bypass to establish a technician session on an internet-facing SimpleHelp server. From there, they deploy TaskWeaver — a JavaScript-based malware loader — which fingerprints the compromised device and communicates with command-and-control infrastructure to receive additional modules.
Then Djinn Stealer does its work. In a single pass, it grabs everything it can find: cloud provider credentials, Git config, SSH keys, Docker credentials, infrastructure-as-code tools like Terraform and Pulumi, secrets management solutions like HashiCorp Vault, and package manager credentials for npm, Cargo, pip, and more.
It also targets AI development tooling specifically — stealing Model Context Protocol (MCP) configurations for Claude, Gemini, Codex, Cline, OpenCode, and Kilo. Cryptocurrency wallets get swept up too, along with browser data, shell history, PGP keys, and database client configurations.
On Linux, the malware goes further, reading /proc virtual files to extract secrets from running processes — API keys, credentials, session tokens, file paths, URLs.
The compromised RMM platform gives attackers a trusted administrative channel for transferring files and executing commands on every system managed through that server. If you’re running SimpleHelp with OIDC authentication, patch immediately.
