Vulnerability database · updated every 15 minutes
New vulnerabilities in the software websites run.
CVEs and security advisories for WordPress, its plugins and themes, npm, PyPI and the servers behind websites, merged from NVD, the CVE Program, GitHub, OSV.dev and WPVulnerability, with CISA's list of the ones attackers exploit.
- Tracked
- 55,384
- Published in 7 days
- 1,885
- Critical in 30 days
- 489
- Actively exploited
- 542
Results for “vLLM”
169 vulnerabilities- Medium 6.9
CVE-2026-100647 · vLLM
vLLM DoS via unbounded `cache_salt` length: multi-hundred-MB salt stalls the single EngineCore scheduler thread (CPU exhaustion)
Published 12 Sept 2026 · Fixed in 0.29.0 · NVD, Project advisories (GitHub) - High 8.7
CVE-2026-93592 · vLLM
Denial of service: a negative token id in `/v1/embeddings` or `/pooling` input kills the vLLM engine via a CUDA device-side assertion
Published 3 Sept 2026 · Fixed in 0.28.0 · NVD, Project advisories (GitHub) - Medium 6.5
CVE-2026-69147 · vLLM
Request-selected PyNvVideoCodec GPU decode bypasses static VRAM reservation
Published 28 Aug 2026 · Fixed in 0.28.0 · NVD, Project advisories (GitHub) - High 8.5
CVE-2026-90553 · vLLM
LlavaOnevision2 processor loader executes attacker model code with `trust_remote_code=False` (inert `trust_remote_code` kwarg to `transformers.get_class_from_dynamic_module`) — RCE from a malicious model
Published 28 Aug 2026 · Fixed in 0.28.0 · NVD, Project advisories (GitHub) - Medium 6.9
CVE-2026-90554 · vLLM
Denial of Service NanoNemoTronVL Video Audio Extraction Bomb
Published 28 Aug 2026 · Fixed in 0.28.0 · NVD, Project advisories (GitHub) - High 7.1
CVE-2026-90555 · vLLM
Speech-to-text audio decode duration limit bypass via forged header sample rate
Published 28 Aug 2026 · Fixed in 0.28.0 · NVD, Project advisories (GitHub) - High 7.5
CVE-2026-37237 · vLLM
vLLM up to and including 0.17.0 allows remote attackers to cause a Denial of Service via memory exhaustion. The AsyncMediaIO.fetch_audio and AsyncMediaIO.fetch_image functions in multimodal/inputs.py fetch user-supplied media URLs using aiohttp and call r.read() without enforcing a maximum response size, allowing an attacker to exhaust server memory by providing a URL to an arbitrarily large file.
Published 28 Aug 2026 · NVD - Medium 6.5
CVE-2026-57173 · vLLM
Unauthenticated audio decompression-bomb DoS in /v1/chat/completions: VLLM_MAX_AUDIO_DECODE_DURATION_S guard not wired into the chat audio path (sibling of CVE-2026-5497)
Published 23 Aug 2026 · Fixed in 0.24.0 · NVD, Project advisories (GitHub) - Medium 6.5
CVE-2026-73560 · vLLM
SSRF + arbitrary local file read in MiMoV2OmniMultiModalProcessor `_fetch_image` and audio loader bypass MediaConnector protections
Published 13 Aug 2026 · Fixed in 0.26.0, 0.26 · NVD, Project advisories (GitHub) - Medium 6.9
CVE-2026-78684 · vLLM
vLLM Unauthenticated Requests Exploit DeepStream Backend Confusion for DoS
Published 11 Aug 2026 · Fixed in 0.27.0 · NVD, Project advisories (GitHub) - Medium 5.3
CVE-2026-73558 · vLLM
Cross-User Data Leak Vulnerability
Published 11 Aug 2026 · Fixed in 0.21.0, 0.27.0 · NVD, Project advisories (GitHub) - Medium 6.3
CVE-2026-73557 · vLLM
Incomplete CVE-2025-62164 remediation can be bypassed by concurrent prompt parts
Published 27 Jul 2026 · Fixed in 0.26.0 · NVD, Project advisories (GitHub) - Medium 5.3
CVE-2026-73556 · vLLM
ReDoS via structured_outputs.regex in the lm-format-enforcer backend (no compile timeout) — missed sibling of GHSA-rwxx-mrjm-wc2m
Published 27 Jul 2026 · Fixed in 0.26.0 · NVD, Project advisories (GitHub) - Medium 5.3
CVE-2026-73555 · vLLM
Unauthenticated Internal Path and Username Disclosure via Validation Error Messages
Published 27 Jul 2026 · Fixed in 0.26.0 · NVD, Project advisories (GitHub) - Medium 4.3
CVE-2026-71486 · vLLM
Derender endpoints decode caller-supplied GenerateResponse token IDs without output bounds
Published 27 Jul 2026 · Fixed in 0.26.0 · NVD, Project advisories (GitHub) - Medium 6.5
CVE-2026-73559 · vLLM
Completion prompt lists fan out into unbounded engine requests
Published 27 Jul 2026 · Fixed in 0.26.0 · NVD, Project advisories (GitHub) - High 7.5
CVE-2026-54234 · vLLM
vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. Prior to 0.24.0, a frontend-legal multi-request speculative decoding workload can cause the rejection sampler to produce a recovered token equal to the model vocabulary size boundary value, which is then converted to negative one when the engine selects the next live token for a request and is written back into…
Published 6 Jul 2026 · Fixed in 0.24.0 · NVD - Medium 6.5
CVE-2026-55646 · vLLM
Speech-to-text upload size limit is enforced after full UploadFile read
Published 2 Jul 2026 · Fixed in 0.24.0 · NVD, Project advisories (GitHub) - High 8.7
CVE-2026-55574 · vLLM
ReDoS via structured_outputs.regex compiled without timeout in xgrammar and outlines backends
Published 2 Jul 2026 · Fixed in 0.24.0 · NVD, Project advisories (GitHub) - High 7.1
CVE-2026-55514 · vLLM
DoS caused by sending `/v1/completions` with prompt embeds payload with models that use M-RoPE
Published 2 Jul 2026 · Fixed in 0.24.0 · NVD, Project advisories (GitHub) - High 8.7
CVE-2026-56340 · vLLM
vLLM versions >= 0.10.2 and < 0.13.0 are missing sparse tensor validation in multimodal embeddings processing. Because PyTorch disables sparse tensor invariant checks by default, an attacker can submit crafted embedding requests with malformed (negative or out-of-bounds) tensor indices, when the prompt-embeds feature is enabled, to trigger crashes or resource exhaustion (denial of service), with…
Published 20 Jun 2026 · Fixed in 0.13.0 · NVD - Medium 5.3
CVE-2025-71379 · vLLM
vLLM versions >= 0.6.3 and < 0.9.0 contain multiple regular expression denial of service (ReDoS) vulnerabilities. Several regex patterns — in vllm/lora/utils.py, the phi4mini tool parser, and the OpenAI-compatible serving chat endpoint — are susceptible to catastrophic backtracking. An attacker submitting crafted input with nested or repeated structures can trigger severe CPU consumption and…
Published 20 Jun 2026 · Fixed in 0.9.0 · NVD - High 7.5
CVE-2026-41523 · vLLM
Security Check Bypass via assert Statement in Activation Function Loading Allows Arbitrary Code Execution
Published 14 Jun 2026 · Fixed in 0.22.0 · NVD, Project advisories (GitHub) - Medium 6.5
CVE-2026-54233 · vLLM
OOM Denial of Service via Audio Decompression Bomb
Published 11 Jun 2026 · Fixed in 0.23.1, 0.24.0 · NVD, Project advisories (GitHub) - Medium 5.3
CVE-2026-54236 · vLLM
vLLM: incomplete CVE-2026-22778 fix leaks PIL repr addresses via Anthropic router (CWE-532)
Published 11 Jun 2026 · Fixed in 0.23.1, 0.24.0 · NVD, Project advisories (GitHub) - Medium 5.3
CVE-2026-53923 · vLLM
GGUF dequantize kernel int truncation exposes uninitialized GPU memory in multi-tenant serving
Published 11 Jun 2026 · Fixed in 0.23.1, 0.24.0 · NVD, Project advisories (GitHub) - Medium 4.8
CVE-2026-12491 · vLLM
vLLM image EXIF Rotation & PNG tRNS Transparency Not Normalized, Causing Mismatch Between Model Input and Expectations
Published 11 Jun 2026 · Fixed in 0.24.0 · Project advisories (GitHub) - Medium 6.9
CVE-2026-54235 · vLLM
temperature=NaN and temperature=Infinity bypass validation and propagate to GPU kernels
Published 11 Jun 2026 · Fixed in 0.23.1, 0.24.0 · NVD, Project advisories (GitHub) - High 7.5
CVE-2026-5497 · vLLM
vLLM versions 0.8.0 and later are vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.load_base64()` method. When processing `video/jpeg` data URLs, the method splits the base64 data string on commas to extract individual JPEG frames without enforcing a frame count limit. An attacker can exploit this by crafting a…
Published 11 Jun 2026 · Fixed in 0.19.0 · NVD - Medium 6.5
CVE-2026-47155 · vLLM
Artifact Pin Decay in vLLM allows pinned deployments to load unpinned code, weights, and processors
Published 10 Jun 2026 · Fixed in 0.22.0 · NVD, Project advisories (GitHub) - High 8.8
CVE-2026-54232 · vLLM
Dependency Confusion Vulnerability in vLLM Dockerfile
Published 9 Jun 2026 · Fixed in 0.22.1 · NVD, Project advisories (GitHub) - Critical 9.1
CVE-2026-48746 · vLLM
OpenAI API Auth Bypass
Published 2 Jun 2026 · Fixed in 0.22.0 · NVD, Project advisories (GitHub) - High 7.5
CVE-2026-44222 · vLLM
vLLM is an inference and serving engine for large language models (LLMs). From 0.6.1 to before 0.20.0, there is a a Token Injection vulnerability in vLLM’s multimodal processing. Unauthenticated, text-only prompts that spell special tokens are interpreted as control. Image and video placeholder sequences supplied without matching data cause vLLM to index into empty grids during input-position…
Published 12 May 2026 · Fixed in 0.20.0 · NVD - Medium 6.5
CVE-2026-44223 · vLLM
extract_hidden_states speculative decoding crashes server on any request with penalty parameters
Published 28 Apr 2026 · Fixed in 0.20.0 · NVD, Project advisories (GitHub) - Low 2.9
CVE-2026-7141 · vLLM
A vulnerability was found in vLLM up to 0.19.0. The affected element is the function has_mamba_layers of the file vllm/v1/kv_cache_interface.py of the component KV Block Handler. Performing a manipulation results in uninitialized resource. It is possible to initiate the attack remotely. The attack is considered to have high complexity. The exploitability is described as difficult. The exploit…
Published 27 Apr 2026 · NVD - Medium 5.4
CVE-2026-34753 · vLLM
Server-Side Request Forgery (SSRF) in `download_bytes_from_url `
Published 3 Apr 2026 · Fixed in 0.19.0 · NVD, Project advisories (GitHub) - Medium 6.5
CVE-2026-34755 · vLLM
Denial of Service via Unbounded Frame Count in video/jpeg Base64 Processing
Published 3 Apr 2026 · Fixed in 0.19.0 · NVD, Project advisories (GitHub) - Medium 6.5
CVE-2026-34756 · vLLM
OOM Denial of Service via Unbounded `n` Parameter in OpenAI API Server
Published 3 Apr 2026 · Fixed in 0.19.0 · NVD, Project advisories (GitHub) - High 7.1
CVE-2026-34760 · vLLM
Downmix Implementation Differences as Attack Vectors Against Audio AI Models
Published 30 Mar 2026 · Fixed in 0.18.0 · NVD, Project advisories (GitHub) - High 8.8
CVE-2026-27893 · vLLM
Hardcoded trust_remote_code=True in NemotronVL and KimiK25 bypasses user security opt-out
Published 26 Mar 2026 · Fixed in 0.18.0 · NVD, Project advisories (GitHub) - Critical 9.8
CVE-2026-25960 · vLLM
SSRF Protection Bypass in vLLM
Published 9 Mar 2026 · Fixed in 0.17.0 · NVD, Project advisories (GitHub) - Critical 9.8
CVE-2026-22778 · vLLM
vLLM RCE In Video Processing
Published 2 Feb 2026 · Fixed in 0.14.1 · NVD, Project advisories (GitHub) - High 7.1
CVE-2026-24779 · vLLM
Server-Side Request Forgery (SSRF) in `MediaConnector`
Published 27 Jan 2026 · Fixed in 0.14.1 · NVD, Project advisories (GitHub) - Critical 9.8
CVE-2026-22807 · vLLM
RCE via auto_map dynamic module loading during model initialization
Published 21 Jan 2026 · Fixed in 0.14.0 · NVD, Project advisories (GitHub) - High 7.5
CVE-2026-22773 · vLLM
DoS in Idefics3 vision models via image payload with ambiguous dimensions
Published 9 Jan 2026 · Fixed in 0.12.0 · NVD, Project advisories (GitHub) - High 8.8
GHSA-mcmc-2m55-j8jj · vLLM
Missing validation of multimodal embeddings leading to DoS and potential RCE
Published 8 Jan 2026 · Fixed in 0.11.1 · Project advisories (GitHub) - High 8.8
CVE-2025-66448 · vLLM
Remote code execution via transformers_utils/get_config
Published 1 Dec 2025 · Fixed in 0.11.1 · NVD, Project advisories (GitHub) - Medium 6.5
CVE-2025-62426 · vLLM
DoS via large Chat Completion or Tokenization requests with specially crafted `chat_template_kwargs`
Published 20 Nov 2025 · Fixed in 0.11.1 · NVD, Project advisories (GitHub) - High 8.8
CVE-2025-62164 · vLLM
prompt_embeds deserialization allows DoS and potential RCE
Published 20 Nov 2025 · Fixed in 0.11.1 · NVD, Project advisories (GitHub) - High 8.3
CVE-2025-62372 · vLLM
DoS with incorrect shape of multimodal embedding inputs
Published 20 Nov 2025 · Fixed in 0.11.1 · NVD, Project advisories (GitHub)
Where it comes from
The list holds every advisory UpHeron reads for the software it watches: WordPress and every plugin and theme its plugin reports, the packages in monitored repositories and manifests, and the servers and frameworks found on monitored websites. One vulnerability from several sources is one entry, under its CVE.
- NVD, the US National Vulnerability Database. This product uses the NVD API but is not endorsed or certified by the NVD.
- New and changed records of the CVE Program, read every 15 minutes, so a rating or fix added after a CVE is published reaches the list, and its alerts, within the hour. CVE content © 1999–2026 The MITRE Corporation, used under the CVE terms of use; CVE® is a registered trademark of The MITRE Corporation.
- OSV.dev, with the GitHub Advisory Database (CC BY 4.0).
- WPVulnerability (CC0) for WordPress plugins and themes, and WordPress's own security advisories.
- Node.js's own security release posts, read every 15 minutes, which name its vulnerabilities the day they are fixed, days before their CVE records.
- The security advisories projects publish on GitHub, and the CVEs libraries such as Brotli name in their release notes, which also count for the runtimes that bundle them.
- CISA's Known Exploited Vulnerabilities catalog.
Each day's new ones, grouped, and every critical one as soon as we find it: on the blog.
Run WordPress? The UpHeron plugin alerts you the day one of these affects your own site.