CVE-2026-34755: CWE-770: Allocation of Resources Without Limits or Throttling in vllm-project vllm
CVE-2026-34755 is a medium severity vulnerability in the vllm-project's vllm inference engine for large language models. Versions from 0. 7. 0 up to but not including 0. 19. 0 contain a flaw in the VideoMediaIO. load_base64() method, which processes video/jpeg base64 data URLs by splitting on commas to extract JPEG frames without enforcing a frame count limit. This bypasses the intended limit on the number of frames, allowing an attacker to send a request with thousands of frames. The server attempts to decode all frames into memory, leading to an out-of-memory (OOM) crash. The issue is fixed in version 0.
AI Analysis
Technical Summary
The vulnerability arises from the load_base64() method in vllm versions >=0.7.0 and <0.19.0, where the num_frames parameter limiting the number of decoded JPEG frames is bypassed when processing video/jpeg base64 data URLs. An attacker can exploit this by sending a single API request containing a large number of comma-separated base64-encoded JPEG frames, causing excessive memory allocation and a denial of service via server crash. This is classified as CWE-770: Allocation of Resources Without Limits or Throttling. The vulnerability has a CVSS v3.1 score of 6.5 (medium severity) with network attack vector, low attack complexity, requiring low privileges and no user interaction, impacting availability only. The issue is resolved in vllm version 0.19.0.
Potential Impact
Successful exploitation results in a denial of service condition due to the server exhausting memory resources and crashing. There is no impact on confidentiality or integrity reported. The attack requires network access and low privileges but no user interaction. No known exploits in the wild have been reported.
Mitigation Recommendations
This vulnerability is fixed in vllm version 0.19.0. Users should upgrade to vllm 0.19.0 or later to remediate this issue. Since the product is not a cloud service, remediation depends on applying this update. Patch status is confirmed by the vendor's versioning information. No additional mitigations are specified.
CVE-2026-34755: CWE-770: Allocation of Resources Without Limits or Throttling in vllm-project vllm
Description
CVE-2026-34755 is a medium severity vulnerability in the vllm-project's vllm inference engine for large language models. Versions from 0. 7. 0 up to but not including 0. 19. 0 contain a flaw in the VideoMediaIO. load_base64() method, which processes video/jpeg base64 data URLs by splitting on commas to extract JPEG frames without enforcing a frame count limit. This bypasses the intended limit on the number of frames, allowing an attacker to send a request with thousands of frames. The server attempts to decode all frames into memory, leading to an out-of-memory (OOM) crash. The issue is fixed in version 0.
AI-Powered Analysis
Machine-generated threat intelligence
Technical Analysis
The vulnerability arises from the load_base64() method in vllm versions >=0.7.0 and <0.19.0, where the num_frames parameter limiting the number of decoded JPEG frames is bypassed when processing video/jpeg base64 data URLs. An attacker can exploit this by sending a single API request containing a large number of comma-separated base64-encoded JPEG frames, causing excessive memory allocation and a denial of service via server crash. This is classified as CWE-770: Allocation of Resources Without Limits or Throttling. The vulnerability has a CVSS v3.1 score of 6.5 (medium severity) with network attack vector, low attack complexity, requiring low privileges and no user interaction, impacting availability only. The issue is resolved in vllm version 0.19.0.
Potential Impact
Successful exploitation results in a denial of service condition due to the server exhausting memory resources and crashing. There is no impact on confidentiality or integrity reported. The attack requires network access and low privileges but no user interaction. No known exploits in the wild have been reported.
Mitigation Recommendations
This vulnerability is fixed in vllm version 0.19.0. Users should upgrade to vllm 0.19.0 or later to remediate this issue. Since the product is not a cloud service, remediation depends on applying this update. Patch status is confirmed by the vendor's versioning information. No additional mitigations are specified.
Technical Details
- Data Version
- 5.2
- Assigner Short Name
- GitHub_M
- Date Reserved
- 2026-03-30T19:17:10.225Z
- Cvss Version
- 3.1
- State
- PUBLISHED
- Remediation Level
- null
Threat ID: 69d402cf0a160ebd92d2b84e
Added to database: 4/6/2026, 7:00:31 PM
Last enriched: 4/6/2026, 7:15:35 PM
Last updated: 4/6/2026, 8:11:56 PM
Views: 4
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.
External Links
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.