Researchers at Black Lotus Labs have identified a new threat known as the PoeLLM botnet. This malware utilizes a poem hosted on GitHub to manage its command-and-control operations.
Using GitHub poems to mask command signals
Black Lotus Labs researchers have uncovered a highly adaptive method of server control that avoids traditional detection. Instead of relying on static IP addresses or known malicious domains, the PoeLLM botnet uses a poem published on GitHub to encode its command-and-control (C2) address.
The flexibility of this method is what makes the botnet particularly dangerous. by simply updating the text within the GitHub poem,an attacker can shift the location of the control server. This allows the malware to remain agile and evade security measures that look for consistent communication patterns with a single, unchanging server.
The 3,400 servers targeted by PoeLLM
The scale of the PoeLLM infection is significant, with the report stating that more than 3,400 servers have already been compromised. These hijacked systems are primarily being used to conduct cryptocurrency mining and facilitate further network exploitation, draining resources from the host machines.
As researchers at Black Lotus Labs discovered, the malware's primary goal is to turn these servers into a distributed workforce for crypto-mining. Beyond the immediate theft of computing power, the botnet also serves as a foothold for attackers to move laterally through compromised networks, potentially leading to much larger data breaches.
Exploiting vulnerabilities in Ollama and LiteLLM
This discovery highlights a growing vulnerability within the rapidly expanding AI software supply chain. The PoeLLM botnet specifically targets machines running outdated versions of open-source AI and Large Language Model (LLM) tools, such as Ollama and LiteLLM.
As organizations rush to integrate generative AI into their workflows, many are inadvertently leaving the door open to attackers by failing to maintain strict patch management.. According to the findings from Black Lotus Labs, the trend of using popular, open-source AI frameworks without rigorous security auditing has created a fertile ground for botnets to thrive.
The missing identity of the PoeLLM mastermind
Despite the technical details revealed by Black Lotus Labs, several critical pieces of the puzzle remain unverified. It is currently unknown who is behind the creation and management of the PoeLLM botnet, or what their specific long-term strategic objectives might be.
Furthermore, the report does not clarify how long this botnet has been operating undetected or if there are other, more sophisticated command-and-control mechanisms being used alongside the GitHub poem.. Until the identity of the operator is established, the full scope of the threat remains difficult to gauge.
Comments 0