Denial of Service Vulnerability in vLLM by the vLLM Project
CVE-2025-48956

7.5HIGH

Key Information:

Vendor

Vllm-project

Status
Vllm
Vendor
CVE Published:
21 August 2025

What is CVE-2025-48956?

vLLM, an inference and serving engine for large language models, is exposed to a Denial of Service vulnerability due to insufficient handling of oversized HTTP headers. An attacker can exploit this risk by sending a single, specially crafted HTTP GET request containing an excessively large header. This vulnerability can cause server memory exhaustion, resulting in a crash or unavailability of the service. As the attack mechanism does not require user authentication, it poses a significant threat, making any remote user capable of triggering this failure. The issue has been addressed in the release version 0.10.1.1.

Affected Version(s)

vllm >= 0.1.0, < 0.10.1.1

References

https://github.com/vllm-project/vllm/security/advisories/...
x_refsource_CONFIRM
https://github.com/vllm-project/vllm/pull/23267
x_refsource_MISC
https://github.com/vllm-project/vllm/commit/d8b736f913a59...
x_refsource_MISC

CVSS V3.1

Score:
7.5
Severity:
HIGH
Confidentiality:
None
Integrity:
None
Availability:
None
Attack Vector:
Network
Attack Complexity:
Low
Privileges Required:
None
User Interaction:
None
Scope:
Unchanged

Timeline

  • Vulnerability published

  • Vulnerability Reserved

.
CVE-2025-48956 : Denial of Service Vulnerability in vLLM by the vLLM Project