vLLM up to 0.7.x outlines_logits_processors.py extra_body allocation of resources

Summaryinfo

A vulnerability was found in vLLM up to 0.7.x. It has been classified as critical. This issue affects some unknown processing of the file vllm/model_executor/guided_decoding/outlines_logits_processors.py. The manipulation of the argument extra_body leads to allocation of resources. This vulnerability is referenced as CVE-2025-29770. Remote exploitation of the attack is possible. No exploit is available. Upgrading the affected component is recommended.

Detailsinfo

A vulnerability was found in vLLM up to 0.7.x. It has been rated as critical. This issue affects an unknown code block of the file vllm/model_executor/guided_decoding/outlines_logits_processors.py. The manipulation of the argument extra_body with an unknown input leads to a allocation of resources vulnerability. Using CWE to declare the problem leads to CWE-770. The product allocates a reusable resource or group of resources on behalf of an actor without imposing any restrictions on the size or number of resources that can be allocated, in violation of the intended security policy for that actor. Impacted is availability. The summary by CVE is:

vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. The outlines library is one of the backends used by vLLM to support structured output (a.k.a. guided decoding). Outlines provides an optional cache for its compiled grammars on the local filesystem. This cache has been on by default in vLLM. Outlines is also available by default through the OpenAI compatible API server. The affected code in vLLM is vllm/model_executor/guided_decoding/outlines_logits_processors.py, which unconditionally uses the cache from outlines. A malicious user can send a stream of very short decoding requests with unique schemas, resulting in an addition to the cache for each request. This can result in a Denial of Service if the filesystem runs out of space. Note that even if vLLM was configured to use a different backend by default, it is still possible to choose outlines on a per-request basis using the guided_decoding_backend key of the extra_body field of the request. This issue applies only to the V0 engine and is fixed in 0.8.0.

It is possible to read the advisory at github.com. The identification of this vulnerability is CVE-2025-29770 since 03/11/2025. The exploitation is known to be easy. The attack may be initiated remotely. Technical details of the vulnerability are known, but there is no available exploit. The attack technique deployed by this issue is T1499 according to MITRE ATT&CK.

Upgrading to version 0.8.0 eliminates this vulnerability.

Be aware that VulDB is the high quality source for vulnerability data.

Productinfo

Name

• vLLM

Version

• 0.1
• 0.2
• 0.3
• 0.4
• 0.5
• 0.6
• 0.7

CPE 2.3info

• 🔍
• 🔍
• 🔍

CPE 2.2info

• 🔍
• 🔍
• 🔍

CVSSv4info

CVSSv3info

CVSSv2info

Exploitinginfo

Threat Intelligenceinfo

Countermeasuresinfo

Timelineinfo

Sourcesinfo

Entryinfo

Be aware that VulDB is the high quality source for vulnerability data.

Discussion