<?xml version='1.0' encoding='UTF-8'?>
<?xml-stylesheet href="/static/style.xsl" type="text/xsl"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/" version="2.0">
  <channel>
    <title>Most recent entries from all</title>
    <link>https://cve.radiocsirt.org</link>
    <description>Contains only the most 10 recent entries.</description>
    <docs>http://www.rssboard.org/rss-specification</docs>
    <generator>python-feedgen</generator>
    <language>en</language>
    <lastBuildDate>Mon, 16 Nov 2026 20:56:37 +0000</lastBuildDate>
    <item>
      <title>EUVD-2026-393903</title>
      <link>https://cve.radiocsirt.org/vuln/euvd-2026-393903</link>
      <description>EUVD-2026-393903</description>
      <guid isPermaLink="false">https://cve.radiocsirt.org/vuln/euvd-2026-393903</guid>
    </item>
    <item>
      <title>fkie_cve-2026-114053</title>
      <link>https://cve.radiocsirt.org/vuln/fkie_cve-2026-114053</link>
      <description>&lt;p&gt;vLLM from 0.7.3 before 0.31.0 forwards the request-supplied mm_processor_kwargs.code_revision to Transformers AutoProcessor.from_pretrained, letting clients select which revision of a model&amp;#39;s dynamic processor code is imported when trust_remote_code is enabled. An attacker controlling a revision in the served model repository can send a multimodal request to the unauthenticated /tokenize endpoint to execute arbitrary Python in the vLLM API process.&lt;/p&gt;</description>
      <content:encoded>&lt;p&gt;vLLM from 0.7.3 before 0.31.0 forwards the request-supplied mm_processor_kwargs.code_revision to Transformers AutoProcessor.from_pretrained, letting clients select which revision of a model&amp;#39;s dynamic processor code is imported when trust_remote_code is enabled. An attacker controlling a revision in the served model repository can send a multimodal request to the unauthenticated /tokenize endpoint to execute arbitrary Python in the vLLM API process.&lt;/p&gt;</content:encoded>
      <guid isPermaLink="false">https://cve.radiocsirt.org/vuln/fkie_cve-2026-114053</guid>
    </item>
    <item>
      <title>GHSA-h3rc-6mm3-gc2m — vLLM: Request-controlled mm_processor_kwargs.code_revision allows remote code execution</title>
      <link>https://cve.radiocsirt.org/vuln/ghsa-h3rc-6mm3-gc2m</link>
      <description>&lt;p&gt;&lt;strong&gt;Affected:&lt;/strong&gt; PyPI: vllm&lt;/p&gt;
&lt;p&gt;## Summary&lt;/p&gt;
&lt;p&gt;An untrusted multimodal request can set `mm_processor_kwargs.code_revision`. When vLLM serves a model whose Hugging Face processor uses dynamic `auto_map` code and the operator enabled `--trust-remote-code`, this request value reaches `AutoProcessor.from_pretrained` and selects the Python revision imported by Transformers.&lt;/p&gt;
&lt;p&gt;This lets a client replace the operator-reviewed processor module with code from another revision of the same model repository. The operator-selected model/data `revision` remains pinned; only the executed processor code changes.&lt;/p&gt;
&lt;p&gt;Importing a Python module is arbitrary code execution in the vLLM API/renderer process. The PoC below demonstrates this with an import-time filesystem write. It is fully offline and does not require a GPU, model weights, or changes to a real Hub repository.&lt;/p&gt;
&lt;p&gt;## Affected snapshot&lt;/p&gt;
&lt;p&gt;- Repository: [`vllm-project/vllm`](https://github.com/vllm-project/vllm)
- Verified commit: [`570735520942b6428ae98f1eb442c6d88adae55e`](https://github.com/vllm-project/vllm/commit/570735520942b6428ae98f1eb442c6d88adae55e)
- Verified with Transformers 5.15.0 and 5.16.1
- Affected release range has not been bisected&lt;/p&gt;
&lt;p&gt;## Affected configuration&lt;/p&gt;
&lt;p&gt;The vulnerable path exists when all of the following are true:&lt;/p&gt;
&lt;p&gt;1. The served model&amp;#39;s processor uses dynamic `auto_map.AutoProcessor` code.
2. The operator enabled `trust_remote_code`.
3. The server can fetch the request-selected revision, or it is already cached.
4. The attacker can create or otherwis…&lt;/p&gt;</description>
      <content:encoded>&lt;p&gt;&lt;strong&gt;Affected:&lt;/strong&gt; PyPI: vllm&lt;/p&gt;
&lt;p&gt;## Summary&lt;/p&gt;
&lt;p&gt;An untrusted multimodal request can set `mm_processor_kwargs.code_revision`. When vLLM serves a model whose Hugging Face processor uses dynamic `auto_map` code and the operator enabled `--trust-remote-code`, this request value reaches `AutoProcessor.from_pretrained` and selects the Python revision imported by Transformers.&lt;/p&gt;
&lt;p&gt;This lets a client replace the operator-reviewed processor module with code from another revision of the same model repository. The operator-selected model/data `revision` remains pinned; only the executed processor code changes.&lt;/p&gt;
&lt;p&gt;Importing a Python module is arbitrary code execution in the vLLM API/renderer process. The PoC below demonstrates this with an import-time filesystem write. It is fully offline and does not require a GPU, model weights, or changes to a real Hub repository.&lt;/p&gt;
&lt;p&gt;## Affected snapshot&lt;/p&gt;
&lt;p&gt;- Repository: [`vllm-project/vllm`](https://github.com/vllm-project/vllm)
- Verified commit: [`570735520942b6428ae98f1eb442c6d88adae55e`](https://github.com/vllm-project/vllm/commit/570735520942b6428ae98f1eb442c6d88adae55e)
- Verified with Transformers 5.15.0 and 5.16.1
- Affected release range has not been bisected&lt;/p&gt;
&lt;p&gt;## Affected configuration&lt;/p&gt;
&lt;p&gt;The vulnerable path exists when all of the following are true:&lt;/p&gt;
&lt;p&gt;1. The served model&amp;#39;s processor uses dynamic `auto_map.AutoProcessor` code.
2. The operator enabled `trust_remote_code`.
3. The server can fetch the request-selected revision, or it is already cached.
4. The attacker can create or otherwis…&lt;/p&gt;</content:encoded>
      <guid isPermaLink="false">https://cve.radiocsirt.org/vuln/ghsa-h3rc-6mm3-gc2m</guid>
    </item>
    <item>
      <title>RHSA-2026:94702 — Red Hat Security Advisory: Red Hat AI Inference 3.5.3 (CPU)</title>
      <link>https://cve.radiocsirt.org/vuln/rhsa-2026:94702</link>
      <description>&lt;p&gt;mcp: mcp: Denial of Service via session memory exhaustion in streamable HTTP servers vllm: vLLM: Denial of Service via Regular Expression processing vllm: vLLM: Denial of Service via negative token ID input vllm: vLLM: Denial of Service via unbounded P2P KV offloading sessions pillow: pillow: Denial of Service via decompression bomb check bypass in multi-frame images pillow: pillow: Denial of Service via decompression bomb bypass in SPIDER images pillow: pillow: Denial of Service via crafted RLE BMP image pillow: Pillow multi-frame loaders can reuse stale native image storage after mode transitions pillow: pillow: Denial of Service via duplicate page references in PDF files pillow: pillow: Denial of Service via integer overflow in text-width calculation pillow: Pillow: Denial of Service via negative archive member size in TarIO pillow: pillow: Denial of Service via repeated EXIF headers pillow: Pillow: Denial of Service via out-of-bounds read in metafile parser pillow: pillow: Arbitrary code execution via JPEG encoder buffer overflow vllm: vLLM: Denial of Service via unbounded media ingestion aiohttp: aiohttp: Information disclosure via unenforced Secure attribute on shared cookies aiohttp: aiohttp: Denial of Service via missing initial request timeout aiohttp: aiohttp: Denial of Service via Max-Age processing overflow aiohttp: aiohttp: Denial of Service via file descriptor exhaustion in multipart request processing aiohttp: aiohttp: Security control bypass via request schem…&lt;/p&gt;</description>
      <content:encoded>&lt;p&gt;mcp: mcp: Denial of Service via session memory exhaustion in streamable HTTP servers vllm: vLLM: Denial of Service via Regular Expression processing vllm: vLLM: Denial of Service via negative token ID input vllm: vLLM: Denial of Service via unbounded P2P KV offloading sessions pillow: pillow: Denial of Service via decompression bomb check bypass in multi-frame images pillow: pillow: Denial of Service via decompression bomb bypass in SPIDER images pillow: pillow: Denial of Service via crafted RLE BMP image pillow: Pillow multi-frame loaders can reuse stale native image storage after mode transitions pillow: pillow: Denial of Service via duplicate page references in PDF files pillow: pillow: Denial of Service via integer overflow in text-width calculation pillow: Pillow: Denial of Service via negative archive member size in TarIO pillow: pillow: Denial of Service via repeated EXIF headers pillow: Pillow: Denial of Service via out-of-bounds read in metafile parser pillow: pillow: Arbitrary code execution via JPEG encoder buffer overflow vllm: vLLM: Denial of Service via unbounded media ingestion aiohttp: aiohttp: Information disclosure via unenforced Secure attribute on shared cookies aiohttp: aiohttp: Denial of Service via missing initial request timeout aiohttp: aiohttp: Denial of Service via Max-Age processing overflow aiohttp: aiohttp: Denial of Service via file descriptor exhaustion in multipart request processing aiohttp: aiohttp: Security control bypass via request schem…&lt;/p&gt;</content:encoded>
      <guid isPermaLink="false">https://cve.radiocsirt.org/vuln/rhsa-2026:94702</guid>
    </item>
    <item>
      <title>WID-SEC-W-2026-4076 — vllm: Mehrere Schwachstellen</title>
      <link>https://cve.radiocsirt.org/vuln/wid-sec-w-2026-4076</link>
      <description>&lt;p&gt;Ein Angreifer kann mehrere Schwachstellen in vllm ausnutzen, um Daten zu manipulieren, einen Denial-of-Service-Angriff durchzuführen oder beliebigen Programmcode auszuführen.&lt;/p&gt;</description>
      <content:encoded>&lt;p&gt;Ein Angreifer kann mehrere Schwachstellen in vllm ausnutzen, um Daten zu manipulieren, einen Denial-of-Service-Angriff durchzuführen oder beliebigen Programmcode auszuführen.&lt;/p&gt;</content:encoded>
      <guid isPermaLink="false">https://cve.radiocsirt.org/vuln/wid-sec-w-2026-4076</guid>
    </item>
  </channel>
</rss>
