Regular Expression Denial of Service in vLLM by VLLM Project
CVE-2025-48887

6.5MEDIUM

Key Information:

Vendor

Vllm-project

Status
Vllm
Vendor
CVE Published:
30 May 2025

What is CVE-2025-48887?

vLLM, an inference and serving engine for large language models, is vulnerable to a Regular Expression Denial of Service (ReDoS). This vulnerability is caused by a complex and nested regular expression used for detecting tool calls, specifically found within the vllm/entrypoints/openai/tool_parsers/pythonic_tool_parser.py file. Attackers can exploit this vulnerability to significantly degrade performance or render the service unavailable due to catastrophic backtracking caused by the regex pattern's multiple nested quantifiers and optional groups. A patch is available in version 0.9.0 to mitigate this issue.

Affected Version(s)

vllm >= 0.6.4, < 0.9.0

References

https://github.com/vllm-project/vllm/security/advisories/...
x_refsource_CONFIRM
https://github.com/vllm-project/vllm/pull/18454
x_refsource_MISC
https://github.com/vllm-project/vllm/commit/4fc1bf813ad80...
x_refsource_MISC

CVSS V3.1

Score:
6.5
Severity:
MEDIUM
Confidentiality:
None
Integrity:
None
Availability:
None
Attack Vector:
Network
Attack Complexity:
Low
Privileges Required:
Low
User Interaction:
None
Scope:
Unchanged

Timeline

  • Vulnerability published

  • Vulnerability Reserved

.