CVE-2026-90553: Improper Control of Generation of Code ('Code Injection') in vllm-project vLLM
vLLM versions before 0.28.0 have a remote code execution vulnerability in the LlavaOnevision2 processor loader. This flaw allows attackers to execute arbitrary code by crafting malicious models that bypass the trust_remote_code parameter, leading to code execution with the privileges of the vLLM process.
AI Analysis
Technical Summary
CVE-2026-90553 is a remote code execution vulnerability in vLLM prior to version 0.28.0. The vulnerability exists because the LlavaOnevision2 processor loader does not respect the trust_remote_code parameter when loading remote processor classes. An attacker can exploit this by providing a malicious model containing arbitrary code in processing_llava_onevision2.py, which executes with the authority of the vLLM process even if trust_remote_code is set to False.
Potential Impact
Successful exploitation allows an attacker to execute arbitrary code remotely with the privileges of the vLLM process. This can lead to full compromise of the affected system or environment running the vulnerable vLLM version.
Mitigation Recommendations
Upgrade to vLLM version 0.28.0 or later where this vulnerability is fixed. No other mitigation details are provided. Patch status is confirmed by the versioning information indicating the fix in 0.28.0.
CVE-2026-90553: Improper Control of Generation of Code ('Code Injection') in vllm-project vLLM
Description
vLLM versions before 0.28.0 have a remote code execution vulnerability in the LlavaOnevision2 processor loader. This flaw allows attackers to execute arbitrary code by crafting malicious models that bypass the trust_remote_code parameter, leading to code execution with the privileges of the vLLM process.
CVSS v4.0
Score 8.5high
Affected software
vllm-project
vLLM
Run on your own infrastructure? Check whether these packages are installed with threat-finder — our free open-source scanner.
AI-Powered Analysis
Machine-generated threat intelligence
Technical Analysis
CVE-2026-90553 is a remote code execution vulnerability in vLLM prior to version 0.28.0. The vulnerability exists because the LlavaOnevision2 processor loader does not respect the trust_remote_code parameter when loading remote processor classes. An attacker can exploit this by providing a malicious model containing arbitrary code in processing_llava_onevision2.py, which executes with the authority of the vLLM process even if trust_remote_code is set to False.
Potential Impact
Successful exploitation allows an attacker to execute arbitrary code remotely with the privileges of the vLLM process. This can lead to full compromise of the affected system or environment running the vulnerable vLLM version.
Mitigation Recommendations
Upgrade to vLLM version 0.28.0 or later where this vulnerability is fixed. No other mitigation details are provided. Patch status is confirmed by the versioning information indicating the fix in 0.28.0.
Technical Details
- Data Version
- 5.2
- Assigner Short Name
- VulnCheck
- Date Reserved
- 2026-09-12T11:13:43.326Z
- Cvss Version
- 4.0
- State
- PUBLISHED
Threat ID: 6aa5465455bf5e2cf54629e9
Added to database: 09/12/2026, 12:32:20 UTC
Last enriched: 09/12/2026, 12:46:31 UTC
Last updated: 09/12/2026, 13:16:32 UTC
Views: 5
Community Reviews
0 reviewsCrowdsource mitigation strategies, share intel context, and vote on the most helpful responses. Sign in to add your voice and help keep defenders ahead.
Want to contribute mitigation steps or threat intel context? Sign in or create an account to join the community discussion.
Actions
Updates to AI analysis require Pro Console access. Upgrade inside Console → Billing.
Need more coverage?
Upgrade to Pro Console for AI refresh and higher limits.
For incident response and remediation, OffSeq services can help resolve threats faster.
Latest Threats
Check if your credentials are on the dark web
Instant breach scanning across billions of leaked records. Free tier available.