CVE-2025-30165: vLLM: pickle RCE in multi-node inference deployments

vLLM is an inference and serving engine for large language models. In a multi-node vLLM deployment using the V0 engine, vLLM uses ZeroMQ for some multi-node communication purposes. The secondary vLLM hosts open a `SUB` ZeroMQ socket and connect to an `XPUB` socket on the primary vLLM host. When data is received on this `SUB` socket, it is deserialized with `pickle`. This is unsafe, as it can be abused to execute code on a remote machine. Since the vulnerability exists in a client that connects to the primary vLLM host, this vulnerability serves as an escalation point. If the primary vLLM host is compromised, this vulnerability could be used to compromise the rest of the hosts in the vLLM deployment. Attackers could also use other means to exploit the vulnerability without requiring access to the primary vLLM host. One example would be the use of ARP cache poisoning to redirect traffic to a malicious endpoint used to deliver a payload with arbitrary code to execute on the target machine. Note that this issue only affects the V0 engine, which has been off by default since v0.8.0. Further, the issue only applies to a deployment using tensor parallelism across multiple hosts, which we do not expect to be a common deployment pattern. Since V0 is has been off by default since v0.8.0 and the fix is fairly invasive, the maintainers of vLLM have decided not to fix this issue. Instead, the maintainers recommend that users ensure their environment is on a secure network in case this pattern is in use. The V1 engine is not affected by this issue.

An attacker positioned on the same network segment as a vLLM multi-node cluster performs ARP cache poisoning to impersonate the primary vLLM host's IP, hijacking the XPUB ZeroMQ socket. Secondary worker nodes — configured to subscribe to the primary's XPUB endpoint — begin receiving attacker-controlled ZeroMQ messages. These messages contain a crafted pickle payload encoding a Python object whose __reduce__ method executes arbitrary OS commands upon deserialization (e.g., reverse shell, credential dump, or ransomware). The attacker achieves simultaneous RCE on all subscribed secondary nodes, gaining control of GPU inference processes, model weights resident in memory, and any secrets accessible to the vLLM service account — all with no authentication and no user interaction required.