Cybersecurity experts have uncovered a series of compromised npm packages that are being used to deliver an AI-enhanced Linux backdoor known as RedC2 4.0. These packages, disguised as calendar and utility tools, integrate a malicious component that activates upon loading, allowing it to execute the payload without additional installation hooks. This makes it particularly insidious, as merely including the package in a project suffices to execute the exploit.
The Trojanized packages appear fully functional, providing the promised utilities while clandestinely embedding a Linux backdoor. This backdoor, RedC2 4.0, communicates with remote servers and facilitates diverse post-exploitation activities. Notably, this variant is marketed on cybercrime platforms as a versatile toolkit for Windows, macOS, and Linux, offering capabilities such as surveillance, credential theft, and payload deployment. It has been actively developed over the past year, with its latest version introducing the RedShell Linux beacon, which is responsible for establishing contact with a command-and-control server.
The RedC2 framework is robust, providing features such as terminal access, file transfer, data collection, and network visualization. It also supports complex operations like in-memory execution and network pivoting. Its AI component, Red Agent, allows operators to use natural language for executing sophisticated tasks, thereby lowering the entry barrier for potential attackers. Despite being marketed as a tool for ethical red teaming, the framework's capabilities highlight the risks associated with AI-driven cyber threats.
This discovery follows a recent supply chain attack involving Rust crates, where malicious dependencies were introduced to execute cross-platform malware. The npm packages' infiltration suggests a likely compromise of the maintainer's credentials, echoing previous attacks linked to North Korean actors.