Veille CVE

Recherche « LMCache »

6 CVE

Critique 3 Élevée 2 Moyenne 1 Faible 0 Non évaluée 0 Exploitées (KEV) Tri : plus récentes « LMCache » ✕
  1. CVE-2026-107207 7.2 Élevée

    LMCache

    LMCache through 0.5.5 contains a server-side request forgery vulnerability in its frontend monitoring service that allows unauthenticated attackers to bypass the proxy allowlist by registering arbitrary hosts. Attackers can add entries via POST /api/proxies and then use /proxy or /proxy2 to reach internal hosts, read responses, and tamper with nodes or stop the heartbeat.

  2. CVE-2026-107206 9.4 Critique

    LMCache

    LMCache through 0.5.5 contains a missing authentication vulnerability in the multiprocess mode HTTP server that allows remote unauthenticated attackers to access management endpoints listening on all interfaces by default. Attackers can read environment credentials via GET /env and configuration via GET /config, clear caches, delete cache objects, and modify tenant quotas to evict other tenants' cached data.

  3. CVE-2026-107205 8.6 Élevée

    LMCache

    LMCache through 0.5.5 contains a missing authentication vulnerability in the multiprocess coordinator that allows remote unauthenticated attackers to access its HTTP fleet control API listening on all interfaces by default. Attackers can register or deregister instances via /instances, overwrite quotas via /quota endpoints, inject events via /events, and enumerate /directory/keys to disrupt caching and disclose placement metadata.

  4. CVE-2026-107204 9.8 Critique

    LMCache

    LMCache through 0.5.5 contains an unauthenticated remote code execution vulnerability that allows remote attackers to execute Python code by posting scripts to the /run_script endpoint. Attackers can recover real builtins through the injected FastAPI app object, bypassing the guarded __import__, to import os and run operating system commands as the LMCache process.

  5. CVE-2026-105192 9.8 Critique

    LMCache

    LMCache multiprocess mode, also called distributed mode, opens an unauthenticated ZeroMQ ROUTER so worker processes can register and share KV cache blocks. Messages on that socket are msgpack. Extension code 1 is passed to DeviceIPCWrapper.Deserialize, which calls pickle.loads, while the server is still decoding request arguments and before the handler runs. A single unauthenticated ZMQ DEALER message to the transport port (default 5555) therefore executes code as the user the LMCache process runs as. Official container images run that process as root. The transport binds to localhost unless the operator sets a routable address with --host, which is how multi-node deployments let peers connect.

  6. CVE-2026-105756 6.5 Moyenne

    vllm-project vllm · Vllm vllm

    vLLM is an inference and serving engine for large language models. Prior to 0.30.0, OpenAI-compatible request models accept a non-empty cache_salt value without enforcing the character and length restrictions required by the IPCCacheServerKey consumer in LMCache-MP. On deployments using the LMCache-MP connector, a salt that contains a forbidden character or exceeds the permitted length can raise an uncaught ValueError during scheduler cache lookup, causing EngineCore to terminate and denying service to all concurrent users. This issue is fixed in version 0.30.0.