JFrog disclosed CVE-2026-105192, a critical flaw rated 9.8 in LMCache, open-source software that speeds up LLM servers such as vLLM. In multiprocess mode the cache runs as a standalone server reached over ZeroMQ with no authentication, and one message type is deserialized with pickle before its type is checked, letting a crafted message run code. The server is exposed only when an operator binds it to a routable address rather than the default localhost, and the code runs with LMCache's privileges, which are root in the project's official container images. No fix exists yet, and JFrog advises keeping the port on localhost or a trusted cluster network.
Lumen Black Lotus Labs detailed PoeLLM, the malware behind the financially motivated Canto Incognito campaign, which has infected more than 3,400 internet-facing servers since April 2026. The actor hides its command-and-control address inside a poem stored in a GitHub repository, changing a few words whenever it rotates infrastructure. It exploits known vulnerabilities in exposed services such as LiteLLM, Gotenberg, Gitea, and Ivanti Sentry, then installs XMRig and Iron miners to abuse victim compute for cryptocurrency mining through a Russian mining service. Compromised hosts become scanners and exploit servers that spread the malware further, and recent traffic suggests experimentation with distributed brute-force attacks.
TrustStrike Labs