LMDeploy vulnerable to arbitrary code execution via eval() of untrusted quant_dtype in model config loading
CVE-2026-33625 · GHSA-3hmm-rh5q-gwwr · published 2026-09-18
- Severity
- High
- CVSS score
- 8.8
- Chance of exploitation (EPSS)
- 0.44%
- Fixed in
- 0.12.3
- Exploitation Public proof of concept
- Automatable No
- Technical impact Total
- CISA SSVC
An attacker who publishes a malicious model on HuggingFace Hub can achieve arbitrary code execution on any machine that loads the model with lmdeploy. This is a supply-chain attack vector affecting all lmdeploy users who load untrusted models. 1. Full remote code execution when loading a malicious model 2. No user interaction beyond running lmdeploy serve or similar with the model 3. Affects all deployment scenarios (local, cloud, production)
Upgrade lmdeploy to 0.12.3 or later.
| Package | Registry | Affected | Fixed in |
|---|---|---|---|
| lmdeploy | PyPI | >= 0.12.1, < 0.12.3 | 0.12.3 |
- HightsParticles: Unbounded HTTP Request Body Buffering in @tsparticles/mcp-server Leads to Remote DoS@tsparticles/mcp-server
- HighMissing request-wide media budget allows single-request frontend resource exhaustionvllm
- HighMCP TypeScript SDK: HTTP server transports and handlers did not limit request body size or JSON-RPC batch length@modelcontextprotocol/hono, @modelcontextprotocol/node, @modelcontextprotocol/sdk, @modelcontextprotocol/server
- HighvLLM : DoS via unvalidated mm_processor_kwargs in multimodal image processors (Isaac, InternVL, H2OVL, etc.)vllm
- HighMCP TypeScript SDK: One HTTP request within the default size limit could exhaust server memory during validation@modelcontextprotocol/sdk, @modelcontextprotocol/server
- HighFrame-count cap bypass in VideoMediaIO.load_base64 for video/jpeg via media_io_kwargs.video.num_framesvllm
- CriticalLMDeploy has Remote Code Execution by Pickle Deserialization via handle_zmq_recv in lmdeploy/lmdeploy/pytorch/disagg/conn/engine_conn.pylmdeploy
- CriticalLMdeploy has Remote Code Execution by Pickle Deserialization via zmq_rpc.call_and_response() in InterLM/lmdeploylmdeploy
- HighLMDeploy has an SSRF bypasslmdeploy
- Highlmdeploy: Hardcoded trust_remote_code=True is an implicit unsafe remote-code load path with no user opt-outlmdeploy
- HighLMDeploy: Arbitrary code execution via hardcoded trust_remote_code=True in lmdeploy model initializationlmdeploy
- HighLMDeploy has Server-Side Request Forgery (SSRF) via Vision-Language Image Loadinglmdeploy
The impact, patch and workaround text comes from the advisory, shortened (GitHub Advisory Database, CC BY 4.0). Proof-of-concept and reproduction sections and code are left out; the original advisory has them.