# Vllm vulnerabilities

Source: https://junglewise.ai/threats/technologies/vllm  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Updated: 12 November 2026, 00:07 UTC. Rebuilt every hour.  
Data: https://junglewise.ai/threats/technologies/vllm.json  
License: CC BY 4.0 with a link back

Junglewise Threat Intelligence has tracked 36 vulnerabilities in Vllm: 0 in the last 7 days and 13 in the last 90 days, 1 of them critical and 0 exploited in the wild. The most recent, CVE-2026-115914, was published on 26 October 2026. 2 vulnerabilities in the last four weeks, down 67% on the four before (6).

## Counts

- Last 7 days: 0
- Last 30 days: 1
- Last 90 days: 13
- All time: 36 (1 critical, 0 exploited in the wild)

## About Vllm

A fast and easy-to-use library for LLM inference and serving.

## Latest Vllm vulnerabilities

- [CVE-2026-115914: vLLM denial of service in DeepSeek-OCR image processing](https://junglewise.ai/threats/cve-2026-115914-vllm-is-an-inference-and-serving-engine-for-large-language) (medium, CVSS 6.5, published 26 Oct 2026)
- [CVE-2026-109311: vLLM heap overflow in audio transcription IAMF parser](https://junglewise.ai/threats/cve-2026-109311-vllm-before-0-29-0-passes-uploaded-audio-to-the-pyav-ffmpeg-av) (medium, CVSS 6.5, published 12 Oct 2026)
- [CVE-2026-105758: vLLM Qwen video backend memory exhaustion via unbounded sampling parameters](https://junglewise.ai/threats/cve-2026-105758-vllm-is-an-inference-and-serving-engine-for-large-language) (medium, CVSS 5.3, published 5 Oct 2026)
- [CVE-2026-105757: vLLM structured-output request handling denial of service](https://junglewise.ai/threats/cve-2026-105757-vllm-is-an-inference-and-serving-engine-for-large-language) (medium, CVSS 6.5, published 5 Oct 2026)
- [CVE-2026-105754: vLLM inference engine input validation bypass in scale-out](https://junglewise.ai/threats/cve-2026-105754-vllm-is-an-inference-and-serving-engine-for-large-language) (medium, CVSS 6.5, published 5 Oct 2026)
- [CVE-2026-105753: vLLM multimodal cache assertion failure and availability impact](https://junglewise.ai/threats/cve-2026-105753-vllm-is-an-inference-and-serving-engine-for-large-language) (medium, CVSS 6.5, published 5 Oct 2026)
- [CVE-2026-93592: vLLM input validation bypass in embeddings and pooling endpoints](https://junglewise.ai/threats/cve-2026-93592-vllm-input-validation-bypass-in-embeddings-and-pooling-endpoints) (high, CVSS 7.5, published 18 Sep 2026)
- [CVE-2026-93436: vLLM memory exhaustion in decode-side metadata cleanup](https://junglewise.ai/threats/cve-2026-93436-vllm-memory-exhaustion-in-decode-side-metadata-cleanup) (high, CVSS 7.5, published 17 Sep 2026)
- [CVE-2026-90555: vLLM audio transcription endpoint denial of service via forged FLAC headers](https://junglewise.ai/threats/cve-2026-90555-vllm-audio-transcription-endpoint-denial-of-service-via-forged) (medium, CVSS 6.5, published 12 Sep 2026)
- [CVE-2026-90554: vLLM NanoNemotronVL audio extraction denial of service](https://junglewise.ai/threats/cve-2026-90554-vllm-nanonemotronvl-audio-extraction-denial-of-service) (medium, CVSS 6.2, published 12 Sep 2026)
- [CVE-2026-90553: vLLM LlavaOnevision2 processor remote code execution via trust_remote_code bypass](https://junglewise.ai/threats/cve-2026-90553-vllm-llavaonevision2-processor-remote-code-execution-via-trust) (high, CVSS 7.8, published 12 Sep 2026)
- [CVE-2026-78684: vLLM DeepStream backend classification and pixel-limit enforcement bypass](https://junglewise.ai/threats/cve-2026-78684-vllm-deepstream-backend-classification-and-pixel-limit) (medium, CVSS 5.3, published 25 Aug 2026)
- [CVE-2026-73560: vLLM SSRF and arbitrary file read in MiMoV2OmniMultiModalProcessor](https://junglewise.ai/threats/cve-2026-73560-vllm-ssrf-and-arbitrary-file-read-in) (medium, CVSS 6.5, published 17 Aug 2026)
- [CVE-2026-73559: vLLM completion prompt lists DoS](https://junglewise.ai/threats/cve-2026-73559-vllm-completion-prompt-lists-dos) (medium, CVSS 6.5, published 13 Aug 2026)
- [CVE-2026-73558: vLLM integer overflow in kernel causing cross-user data leak](https://junglewise.ai/threats/cve-2026-73558-vllm-integer-overflow-in-kernel-causing-cross-user-data-leak) (medium, CVSS 5.3, published 13 Aug 2026)
- [CVE-2026-73557: vLLM concurrent prompt-embedding guard bypass](https://junglewise.ai/threats/cve-2026-73557-vllm-concurrent-prompt-embedding-guard-bypass) (medium, CVSS 5.3, published 13 Aug 2026)
- [CVE-2026-73556: vLLM ReDoS in lm-format-enforcer backend regex parsing](https://junglewise.ai/threats/cve-2026-73556-vllm-redos-in-lm-format-enforcer-backend-regex-parsing) (medium, CVSS 5.3, published 13 Aug 2026)
- [CVE-2026-73555: vLLM information disclosure via validation error messages](https://junglewise.ai/threats/cve-2026-73555-vllm-information-disclosure-via-validation-error-messages) (medium, CVSS 5.3, published 13 Aug 2026)
- [CVE-2025-6242: PYSEC-2026-2011 - vLLM is vulnerable to Server-Side Request Forgery (SSRF) through `MediaConnector` class](https://junglewise.ai/threats/cve-2025-6242-vllm-server-side-request-forgery-in-mediaconnector) (low, CVSS 3.1, published 7 Jul 2026)
- [vLLM missing tensor validation in multimodal embeddings](https://junglewise.ai/threats/vllm-missing-tensor-validation-in-multimodal-embeddings-af4fc3f3) (high, CVSS 8.8, published 20 Jun 2026)
- [vLLM improper image metadata normalization in image processing](https://junglewise.ai/threats/vllm-improper-image-metadata-normalization-in-image-processing-68504b9d) (medium, CVSS 4.8, published 17 Jun 2026)
- [CVE-2026-4944: vLLM remote code execution via hardcoded trust_remote_code parameter](https://junglewise.ai/threats/cve-2026-4944-vllm-remote-code-execution-via-hardcoded-trust-remote-code) (high, CVSS 8.8, published 28 May 2026)
- [CVE-2026-44223: vLLM denial of service in extract_hidden_states speculative decoding](https://junglewise.ai/threats/cve-2026-44223-vllm-denial-of-service-in-extract-hidden-states-speculative) (medium, CVSS 6.5, published 12 May 2026)
- [CVE-2026-7141: vLLM makes Use of Uninitialized Resource](https://junglewise.ai/threats/cve-2026-7141-vllm-makes-use-of-uninitialized-resource) (medium, CVSS 5.6, published 27 Apr 2026)
- [CVE-2026-34756: vLLM Denial of Service via unbounded n parameter in OpenAI API server](https://junglewise.ai/threats/cve-2026-34756-vllm-denial-of-service-via-unbounded-n-parameter-in-openai-api) (medium, CVSS 6.5, published 6 Apr 2026)

## Most severe Vllm vulnerabilities

- [CVE-2026-22778: vLLM remote code execution via JPEG2000 heap overflow and ASLR bypass](https://junglewise.ai/threats/cve-2026-22778-vllm-remote-code-execution-via-jpeg2000-heap-overflow-and-aslr) (critical, CVSS 9.8, published 2 Feb 2026)
- [CVE-2026-27893: vLLM remote code execution via hardcoded remote code trust in models](https://junglewise.ai/threats/cve-2026-27893-vllm-remote-code-execution-via-hardcoded-remote-code-trust-in) (high, CVSS 8.8, published 27 Mar 2026)
- [CVE-2025-62164: vLLM unsafe deserialization in Completions API prompt embeddings](https://junglewise.ai/threats/cve-2025-62164-vllm-unsafe-deserialization-in-completions-api-prompt-embeddings) (high, CVSS 8.8, published 21 Nov 2025)
- [CVE-2026-22807: vLLM arbitrary code execution via auto_map dynamic module loading](https://junglewise.ai/threats/cve-2026-22807-vllm-arbitrary-code-execution-via-auto-map-dynamic-module-loading) (high, CVSS 8.8, published 21 Jan 2026)
- [vLLM missing tensor validation in multimodal embeddings](https://junglewise.ai/threats/vllm-missing-tensor-validation-in-multimodal-embeddings-af4fc3f3) (high, CVSS 8.8, published 20 Jun 2026)
- [CVE-2026-4944: vLLM remote code execution via hardcoded trust_remote_code parameter](https://junglewise.ai/threats/cve-2026-4944-vllm-remote-code-execution-via-hardcoded-trust-remote-code) (high, CVSS 8.8, published 28 May 2026)
- [CVE-2026-90553: vLLM LlavaOnevision2 processor remote code execution via trust_remote_code bypass](https://junglewise.ai/threats/cve-2026-90553-vllm-llavaonevision2-processor-remote-code-execution-via-trust) (high, CVSS 7.8, published 12 Sep 2026)
- [CVE-2026-93436: vLLM memory exhaustion in decode-side metadata cleanup](https://junglewise.ai/threats/cve-2026-93436-vllm-memory-exhaustion-in-decode-side-metadata-cleanup) (high, CVSS 7.5, published 17 Sep 2026)
- [CVE-2026-93592: vLLM input validation bypass in embeddings and pooling endpoints](https://junglewise.ai/threats/cve-2026-93592-vllm-input-validation-bypass-in-embeddings-and-pooling-endpoints) (high, CVSS 7.5, published 18 Sep 2026)
- [CVE-2026-25960: vLLM SSRF protection bypass in MediaConnector via URL parsing inconsistency](https://junglewise.ai/threats/cve-2026-25960-vllm-ssrf-protection-bypass-in-mediaconnector-via-url-parsing) (high, CVSS 7.1, published 9 Mar 2026)

## Vulnerabilities per week

2 vulnerabilities in the last four weeks, down 67% on the four before (6).

| Week of | Vulns | Critical | Exploited |
| --- | --- | --- | --- |
| 17 Aug 2026 | 1 | 0 | 0 |
| 24 Aug 2026 | 1 | 0 | 0 |
| 31 Aug 2026 | 0 | 0 | 0 |
| 7 Sep 2026 | 3 | 0 | 0 |
| 14 Sep 2026 | 2 | 0 | 0 |
| 21 Sep 2026 | 0 | 0 | 0 |
| 28 Sep 2026 | 0 | 0 | 0 |
| 5 Oct 2026 | 4 | 0 | 0 |
| 12 Oct 2026 | 1 | 0 | 0 |
| 19 Oct 2026 | 0 | 0 | 0 |
| 26 Oct 2026 | 1 | 0 | 0 |
| 2 Nov 2026 | 0 | 0 | 0 |
| 9 Nov 2026 | 0 | 0 | 0 |

## Sources

NVD, GitHub Security Advisories, OSV, the CISA Known Exploited Vulnerabilities catalog, FIRST EPSS and vendor advisories. Rebuilt every hour. Cite as: Junglewise Threat Intelligence, "Vllm vulnerabilities", https://junglewise.ai/threats/technologies/vllm, 12 November 2026.