Engine crash via negative token ID
Published Sep 18, 2026 · Updated Sep 18, 2026
Denial of service in vLLM before 0.28.0 allows remote attackers to crash the engine through negative token IDs in API requests. InputProcessor._validate_model_input checks only the maximum token ID, so a negative value reaches the GPU embedding-table lookup and triggers a CUDA assertion. Exposed /v1/embeddings or /pooling endpoints require no authentication; one request poisons the process GPU context and stops all clients until restart.
Summary
What happened
Denial of service in vLLM before 0.28.0 allows remote attackers to crash the engine through negative token IDs in API requests. InputProcessor._validate_model_input checks only the maximum token ID, so a negative value reaches the GPU embedding-table lookup and triggers a CUDA assertion. Exposed /v1/embeddings or /pooling endpoints require no authentication; one request poisons the process GPU context and stops all clients until restart.
The record
- CVE
- CVE-2026-93592
- Published
- Sep 18, 2026
- Updated
- Sep 18, 2026
- Vendor
- vLLM Project
- Product
- vLLM
- Classifications
- CWE-129, T1499.004
- Attack vector
- network
- Privileges
- unauthenticated
Timeline
How it unfolded
- Sep 18, 2026CVE publishedPublication date reported by the CVE source.
- Sep 18, 2026Record updatedLatest update available in the CVE record.
Exploitability
Present is not the same as exploitable
Compare your product and version with the public record. A matching version still requires validation against your environment.
Is a vulnerable build present?
Compare these published version ranges with your installed build and any vendor patches.
- Affected versionversion=0 <0.28.0
What conditions does exploitation require?
What is affected?
Published CVSS scores
CVSS describes severity. EPSS estimates exploitation probability.
Attacks
What attackers are doing with it
Daily unique IPs observed by Shadowserver honeypots for known exploited vulnerabilities (KEVs). Missing observations do not establish an absence of attacks.
Activity data is delayed. The latest report is Sep 28, 2026.
Weakness, pattern, technique
Public exploit references
- Upstream negative token ID proof of conceptproof of concept · demonstrated
Labels summarize the accepted research assessment. They do not indicate a test against your environment.
Technologies
Your stack
See the directory against your own environment.
Your stack
Check the software in your environment
Get started to see how Hinoki identifies affected software and validates exploitability in your environment.
Get started