// For flags

CVE-2025-46570

vLLM’s Chunk-Based Prefix Caching Vulnerable to Potential Timing Side-Channel

Severity Score

2.6
*CVSS v3.1

Exploit Likelihood

*EPSS

Affected Versions

*CPE

Public Exploits

0
*Multiple Sources

Exploited in Wild

-
*KEV

Decision

Track
*SSVC
Descriptions

vLLM is an inference and serving engine for large language models (LLMs). Prior to version 0.9.0, when a new prompt is processed, if the PageAttention mechanism finds a matching prefix chunk, the prefill process speeds up, which is reflected in the TTFT (Time to First Token). These timing differences caused by matching chunks are significant enough to be recognized and exploited. This issue has been patched in version 0.9.0.

*Credits: N/A
CVSS Scores
Attack Vector
Network
Attack Complexity
High
Privileges Required
Low
User Interaction
Required
Scope
Unchanged
Confidentiality
Low
Integrity
None
Availability
None
Attack Vector
Network
Attack Complexity
High
Authentication
Single
Confidentiality
Partial
Integrity
None
Availability
None
* Common Vulnerability Scoring System
SSVC
  • Decision:Track
Exploitation
None
Automatable
No
Tech. Impact
Partial
* Organization's Worst-case Scenario
Timeline
  • 2025-04-24 CVE Reserved
  • 2025-05-29 CVE Published
  • 2025-05-29 CVE Updated
  • 2025-06-04 EPSS Updated
  • ---------- Exploited in Wild
  • ---------- KEV Due Date
  • ---------- First Exploit
CWE
  • CWE-208: Observable Timing Discrepancy
CAPEC
Affected Vendors, Products, and Versions
Vendor Product Version Other Status
Vendor Product Version Other Status <-- --> Vendor Product Version Other Status
Vllm-project
Search vendor "Vllm-project"
 Vllm
Search vendor "Vllm-project" for product "Vllm"
 < 0.9.0
Search vendor "Vllm-project" for product "Vllm" and version " < 0.9.0"
en
Affected