<?xml version="1.0" encoding="UTF-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
  <id>https://stackalarm.dev/news/vllm/</id>
  <title>vllm security news, by stackalarm</title>
  <subtitle>Announced releases, fixes, advisories and alerts about vllm over the last 30 days.</subtitle>
  <link href="https://stackalarm.dev/news/vllm/" />
  <link rel="self" href="https://stackalarm.dev/news/vllm/feed.xml" />
  <updated>2026-11-13T11:32:04Z</updated>
  <author><name>stackalarm</name></author>
  <entry>
    <id>https://wid.cert-bund.de/portal/wid/securityadvisory?name=WID-SEC-2026-4195</id>
    <title>[advisory] [UPDATE] [mittel] vllm: Mehrere Schwachstellen</title>
    <updated>2026-11-13T11:32:04Z</updated>
    <link href="https://wid.cert-bund.de/portal/wid/securityadvisory?name=WID-SEC-2026-4195" />
    <category term="advisory" />
    <summary>Source: cert-bund.</summary>
  </entry>
  <entry>
    <id>https://wid.cert-bund.de/portal/wid/securityadvisory?name=WID-SEC-2026-4076</id>
    <title>[advisory] [UPDATE] [hoch] vllm: Mehrere Schwachstellen</title>
    <updated>2026-11-13T10:12:31Z</updated>
    <link href="https://wid.cert-bund.de/portal/wid/securityadvisory?name=WID-SEC-2026-4076" />
    <category term="advisory" />
    <summary>Source: cert-bund.</summary>
  </entry>
  <entry>
    <id>https://wid.cert-bund.de/portal/wid/securityadvisory?name=WID-SEC-2026-3828</id>
    <title>[advisory] [UPDATE] [mittel] vllm: Mehrere Schwachstellen</title>
    <updated>2026-11-13T10:12:06Z</updated>
    <link href="https://wid.cert-bund.de/portal/wid/securityadvisory?name=WID-SEC-2026-3828" />
    <category term="advisory" />
    <summary>Source: cert-bund.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-fpch-3xrf-7q96</id>
    <title>[advisory] [medium] Thinking-budget sparse state aliases onto a co-batched request during a mixed `SWAP`, forcing the reasoning-end token into another request's output (GHSA-fpch-3xrf-7q96)</title>
    <updated>2026-11-09T17:30:49Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-fpch-3xrf-7q96" />
    <category term="advisory" />
    <summary>Affects vllm &gt;= 0.21.0, &lt; 0.28.0 (fixed in 0.28.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-v97p-wpxp-px84</id>
    <title>[advisory] [medium] Cross-user token data leakage in async scheduling + Pipeline Parallelism (GHSA-v97p-wpxp-px84)</title>
    <updated>2026-11-09T17:29:51Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-v97p-wpxp-px84" />
    <category term="advisory" />
    <summary>Affects vllm &gt;= 0.16.0, &lt; 0.21.0 (fixed in 0.21.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-mwgq-9582-w2vf</id>
    <title>[advisory] [medium] Response amplification in OpenAI validation error handler (GHSA-mwgq-9582-w2vf)</title>
    <updated>2026-11-07T08:04:02Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-mwgq-9582-w2vf" />
    <category term="advisory" />
    <summary>Affects vllm &gt;= 0.3.0, &lt; 0.30.0 (fixed in 0.30.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-q9v9-qpjp-fc24</id>
    <title>[advisory] [high] Unauthenticated EngineCore crash via out-of-range prompt token id under skip_tokenizer_init (incomplete fix of GHSA-25q3-v2hm-8vpf) (GHSA-q9v9-qpjp-fc24)</title>
    <updated>2026-11-02T19:50:12Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-q9v9-qpjp-fc24" />
    <category term="advisory" />
    <summary>Affects vllm &gt;= 0.28.0, &lt; 0.33.0 (fixed in 0.33.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-qrj2-j723-m6w8</id>
    <title>[advisory] [medium] vLLM Speculators Config verifier.name_or_path Element Directory Traversal Information Disclosure Vulnerability (GHSA-qrj2-j723-m6w8)</title>
    <updated>2026-11-02T19:49:32Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-qrj2-j723-m6w8" />
    <category term="advisory" />
    <summary>Affects vllm &lt; 0.33.0 (fixed in 0.33.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-5r3c-5x4v-8gjh</id>
    <title>[advisory] [high] Missing request-wide media budget allows single-request frontend resource exhaustion (GHSA-5r3c-5x4v-8gjh)</title>
    <updated>2026-11-02T19:48:58Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-5r3c-5x4v-8gjh" />
    <category term="advisory" />
    <summary>Affects vllm &gt;= 0.6.0, &lt; 0.33.0 (fixed in 0.33.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-894j-wg8m-63h8</id>
    <title>[advisory] [medium] vLLM OpenCVDynamicOpenPanguVideoBackend: request-level num_frames without a ceiling — unauthenticated memory/CPU exhaustion (OOM-kill) on any video-capable deployment (GHSA-894j-wg8m-63h8)</title>
    <updated>2026-10-27T22:24:06Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-894j-wg8m-63h8" />
    <category term="advisory" />
    <summary>Affects vllm &gt;= 0.30.0, &lt; 0.31.0 (fixed in 0.31.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-w4cg-mhc7-mgw3</id>
    <title>[advisory] [medium] MiniMax M3 video preprocessing permits request-driven CPU resource exhaustion through min_pixels (GHSA-w4cg-mhc7-mgw3)</title>
    <updated>2026-10-27T22:22:44Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-w4cg-mhc7-mgw3" />
    <category term="advisory" />
    <summary>Affects vllm &gt;= 0.24.0, &lt; 0.31.0 (fixed in 0.31.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-vxqj-p4gw-9h4c</id>
    <title>[advisory] [medium] Incomplete CVE-2026-34755 remediation can be bypassed by request-level num_frames (GHSA-vxqj-p4gw-9h4c)</title>
    <updated>2026-10-27T22:22:11Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-vxqj-p4gw-9h4c" />
    <category term="advisory" />
    <summary>Affects vllm &gt;= 0.19.0, &lt; 0.31.0 (fixed in 0.31.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-9g3c-x747-c6wr</id>
    <title>[advisory] [medium] Quadratic Cohere2 Vision preprocessing DoS via request-controlled max_patches (GHSA-9g3c-x747-c6wr)</title>
    <updated>2026-10-27T22:20:43Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-9g3c-x747-c6wr" />
    <category term="advisory" />
    <summary>Affects vllm &gt;= 0.10.1, &lt; 0.31.0 (fixed in 0.31.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-5q78-j82c-2vr2</id>
    <title>[advisory] [medium] vLLM denial of service via prompt embeds on XD-RoPE models (incomplete fix of CVE-2026-55514) (GHSA-5q78-j82c-2vr2)</title>
    <updated>2026-10-27T22:18:41Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-5q78-j82c-2vr2" />
    <category term="advisory" />
    <summary>Affects vllm &gt;= 0.28.0, &lt; 0.30.0 (fixed in 0.30.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-m9fg-w9qh-67pc</id>
    <title>[advisory] [medium] sanitize_message() address regex is lowercase-only and does not match PIL's uppercase __repr__, leaking live heap addresses to unauthenticated clients (GHSA-m9fg-w9qh-67pc)</title>
    <updated>2026-10-27T22:14:15Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-m9fg-w9qh-67pc" />
    <category term="advisory" />
    <summary>Affects vllm &gt;= 0.14.0rc1, &lt; 0.31.0 (fixed in 0.31.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-q97c-p368-75c4</id>
    <title>[advisory] [medium] Out-of-vocabulary prompt token id passes input validation and triggers a CUDA device-side assert (engine crash / DoS) (CVE-2026-115916) [published 2026-10-20, updated 2026-10-27]</title>
    <updated>2026-10-27T07:17:12Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-q97c-p368-75c4" />
    <category term="advisory" />
    <summary>Affects vllm &lt;= 0.31.0 (fixed in 0.32.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-x567-49g5-p4cf</id>
    <title>[advisory] [medium] Responses API bypasses trust_request_chat_template for chat_template_kwargs (CVE-2026-115918) [published 2026-10-20, updated 2026-10-26]</title>
    <updated>2026-10-26T17:21:05Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-x567-49g5-p4cf" />
    <category term="advisory" />
    <summary>Affects vllm &gt;= 0.16.0, &lt;= 0.23.0 (fixed in 0.32.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-vvpg-rrww-wpmh</id>
    <title>[advisory] [low] Permissive default CORS origins allow cross-origin requests to the vLLM HTTP server (CVE-2026-115917) [published 2026-10-20, updated 2026-10-26]</title>
    <updated>2026-10-26T17:19:30Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-vvpg-rrww-wpmh" />
    <category term="advisory" />
    <summary>Affects vllm All versions (default config) (fixed in 0.32.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-h3rc-j82p-2c3h</id>
    <title>[advisory] [high] Untrusted EPS inputs can trigger denial of service or Ghostscript RCE in vLLM (CVE-2026-115915) [published 2026-10-20, updated 2026-10-26]</title>
    <updated>2026-10-26T17:11:34Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-h3rc-j82p-2c3h" />
    <category term="advisory" />
    <summary>Affects vllm 0.30.0 (fixed in 0.32.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-gq5q-89c7-vhvx</id>
    <title>[advisory] [medium] DeepSeek-OCR family: an image whose normalized pixels sum to exactly 0 makes embed_multimodal return None, tripping the encoder-output sanity assert — one unauthenticated request kills the whole vLLM engine (default config) (CVE-2026-115914) [published 2026-10-20, updated 2026-10-26]</title>
    <updated>2026-10-26T17:05:57Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-gq5q-89c7-vhvx" />
    <category term="advisory" />
    <summary>Affects vllm 0.30.0 (fixed in 0.32.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-8397-8qf6-f2gm</id>
    <title>[advisory] [medium] Phi-4 multimodal audio encoder materializes and retains full attention matrices: a single max-length audio OOM-crashes the vLLM engine (CVE-2026-115913) [published 2026-10-20, updated 2026-10-26]</title>
    <updated>2026-10-26T16:56:11Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-8397-8qf6-f2gm" />
    <category term="advisory" />
    <summary>Affects vllm 0.31.0 (fixed in 0.32.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-6757-hggm-hp2w</id>
    <title>[advisory] [high] Public multimodal SHM handle forgery reaches worker-side `pickle.loads()` with insecure serialization disabled (CVE-2026-115912) [published 2026-10-20, updated 2026-10-26]</title>
    <updated>2026-10-26T16:53:06Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-6757-hggm-hp2w" />
    <category term="advisory" />
    <summary>Affects vllm &lt;= 0.25.1 (fixed in 0.32.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-542m-9234-8gvf</id>
    <title>[advisory] [medium] DoS via unbounded chat_template Jinja2 rendering exhausts the renderer thread pool (CVE-2026-115911) [published 2026-10-20, updated 2026-10-26]</title>
    <updated>2026-10-26T16:47:02Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-542m-9234-8gvf" />
    <category term="advisory" />
    <summary>Affects vllm &gt;= 0.19.0 (fixed in 0.32.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-823j-m4cj-hmjf</id>
    <title>[advisory] [high] MOSS-Audio request-controlled processor cache permits remote memory exhaustion (CVE-2026-114052) [published 2026-10-06, updated 2026-10-23]</title>
    <updated>2026-10-23T18:59:09Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-823j-m4cj-hmjf" />
    <category term="advisory" />
    <summary>Affects vllm &gt; 0.24.0, &lt; 0.31.0 (fixed in &gt;= 0.31.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-2823-qmq8-rwvj</id>
    <title>[advisory] [medium] Loose `cache_salt` validation lets a single request kill EngineCore on LMCache-MP deployments — uncaught downstream `ValueError` denial of service (CVE-2026-105756) [published 2026-09-23, updated 2026-10-22]</title>
    <updated>2026-10-22T17:05:44Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-2823-qmq8-rwvj" />
    <category term="advisory" />
    <summary>Affects vllm &lt; 0.30.0 (fixed in &gt;= 0.30.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-p92p-rxj5-7p2x</id>
    <title>[advisory] [medium] Forged multimodal UUID + P0/P1 processor-cache eviction drift: unauthenticated remote crash of the whole vLLM engine (CVE-2026-114050) [published 2026-10-06, updated 2026-10-21]</title>
    <updated>2026-10-21T15:52:19Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-p92p-rxj5-7p2x" />
    <category term="advisory" />
    <summary>Affects vllm &lt; 0.31.0 (fixed in &gt;= 0.31.0). Source: vllm-advisories.</summary>
  </entry>
  <entry>
    <id>https://github.com/vllm-project/vllm/security/advisories/GHSA-frvg-qw88-hww9</id>
    <title>[advisory] [medium] vllm p2p never connected zmq_exhaustion (CVE-2026-94624)</title>
    <updated>2026-10-20T07:35:45Z</updated>
    <link href="https://github.com/vllm-project/vllm/security/advisories/GHSA-frvg-qw88-hww9" />
    <category term="advisory" />
    <summary>Affects vllm 0.26.0 (fixed in 0.32.0). Source: vllm-advisories.</summary>
  </entry>
</feed>
