<?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 21:11:41 +0000</lastBuildDate>
    <item>
      <title>EUVD-2026-390631</title>
      <link>https://cve.radiocsirt.org/vuln/euvd-2026-390631</link>
      <description>EUVD-2026-390631</description>
      <guid isPermaLink="false">https://cve.radiocsirt.org/vuln/euvd-2026-390631</guid>
    </item>
    <item>
      <title>fkie_cve-2026-114050</title>
      <link>https://cve.radiocsirt.org/vuln/fkie_cve-2026-114050</link>
      <description>&lt;p&gt;vLLM before 0.31.0 uses a client-supplied multimodal uuid verbatim as cache identity with no consistency check on cache hits, allowing engine crashes. Unauthenticated attackers can desynchronize the P0/P1 processor caches to trigger a fatal token-placeholder mismatch, killing the entire serving instance until restarted.&lt;/p&gt;</description>
      <content:encoded>&lt;p&gt;vLLM before 0.31.0 uses a client-supplied multimodal uuid verbatim as cache identity with no consistency check on cache hits, allowing engine crashes. Unauthenticated attackers can desynchronize the P0/P1 processor caches to trigger a fatal token-placeholder mismatch, killing the entire serving instance until restarted.&lt;/p&gt;</content:encoded>
      <guid isPermaLink="false">https://cve.radiocsirt.org/vuln/fkie_cve-2026-114050</guid>
    </item>
    <item>
      <title>GHSA-p92p-rxj5-7p2x — vLLM: Forged multimodal UUID + P0/P1 processor-cache eviction drift: unauthenticated remote crash of the whole vLLM eng…</title>
      <link>https://cve.radiocsirt.org/vuln/ghsa-p92p-rxj5-7p2x</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;Multimodal content parts (`image_url`, `video_url`, `input_audio`, ...) accept a user-supplied `uuid` field that vLLM uses **verbatim as the item&amp;#39;s cache identity** (`mm_hash`). The two frontend/engine processor caches (P0 &amp;#34;sender&amp;#34; cache in the API process, P1 &amp;#34;receiver&amp;#34; cache in the EngineCore process) are independent LRU caches whose evictions are **not synchronized**, and the P1 hit path performs **zero consistency validation** — it silently substitutes the cached item&amp;#39;s tensor data for whatever the current request actually contains.&lt;/p&gt;
&lt;p&gt;A single unauthenticated client can exploit this to kill the entire engine, deterministically, with default server flags:&lt;/p&gt;
&lt;p&gt;1. **Seed**: send a small image with `uuid=&amp;#34;U&amp;#34;`. Both caches now map `U` to an item with **N_A** visual tokens (e.g. 4).
2. **Drift**: send ~70 requests, each carrying a *unique* uuid, a max-size image, and an over-long text prompt. Each request is rejected with 400 — but **after** its image has been inserted into the P0 cache (insertion happens during chat rendering, before prompt-length validation), so these requests fill P0 **without ever reaching the engine/P1**. `U` is evicted from P0 only.
3. **Kill**: send a *large* image with `uuid=&amp;#34;U&amp;#34;`. P0 misses → the image is freshly processed and the prompt is spliced with **N_B** (e.g. 2048+) placeholder tokens; the request carries full data to the engine, but P1 still has `U` → the engine **discards the fresh data and substitutes the stale N_A-row tensor**. The…&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;Multimodal content parts (`image_url`, `video_url`, `input_audio`, ...) accept a user-supplied `uuid` field that vLLM uses **verbatim as the item&amp;#39;s cache identity** (`mm_hash`). The two frontend/engine processor caches (P0 &amp;#34;sender&amp;#34; cache in the API process, P1 &amp;#34;receiver&amp;#34; cache in the EngineCore process) are independent LRU caches whose evictions are **not synchronized**, and the P1 hit path performs **zero consistency validation** — it silently substitutes the cached item&amp;#39;s tensor data for whatever the current request actually contains.&lt;/p&gt;
&lt;p&gt;A single unauthenticated client can exploit this to kill the entire engine, deterministically, with default server flags:&lt;/p&gt;
&lt;p&gt;1. **Seed**: send a small image with `uuid=&amp;#34;U&amp;#34;`. Both caches now map `U` to an item with **N_A** visual tokens (e.g. 4).
2. **Drift**: send ~70 requests, each carrying a *unique* uuid, a max-size image, and an over-long text prompt. Each request is rejected with 400 — but **after** its image has been inserted into the P0 cache (insertion happens during chat rendering, before prompt-length validation), so these requests fill P0 **without ever reaching the engine/P1**. `U` is evicted from P0 only.
3. **Kill**: send a *large* image with `uuid=&amp;#34;U&amp;#34;`. P0 misses → the image is freshly processed and the prompt is spliced with **N_B** (e.g. 2048+) placeholder tokens; the request carries full data to the engine, but P1 still has `U` → the engine **discards the fresh data and substitutes the stale N_A-row tensor**. The…&lt;/p&gt;</content:encoded>
      <guid isPermaLink="false">https://cve.radiocsirt.org/vuln/ghsa-p92p-rxj5-7p2x</guid>
    </item>
    <item>
      <title>RHSA-2026:95872 — Red Hat Security Advisory: RHOAI 3.5.3 - Red Hat OpenShift AI</title>
      <link>https://cve.radiocsirt.org/vuln/rhsa-2026:95872</link>
      <description>&lt;p&gt;markdown: python-markdown: Denial of Service via inefficient inline link parsing undici: undici: Information disclosure and Denial of Service via malformed Cache-Control directives @fastify/static: @fastify/static: Information disclosure via route guard bypass fast-uri: fast-uri: Host confusion vulnerability via backslash in URI authority undici: undici: Denial of Service via unrequested WebSocket subprotocol extract-zip: extract-zip: Arbitrary file write via symlink in archive encoding/asn1: golang: Go encoding/asn1: Denial of Service via excessive recursion in Unmarshal find-my-way: find-my-way: Denial of Service vulnerability in HTTP/2 server systeminformation: systeminformation: Arbitrary Command Execution via OS Command Injection mlflow: mlflow: Arbitrary file write via FileStore path traversal mcp: mcp: Denial of Service via session memory exhaustion in streamable HTTP servers nltk: NLTK: Information Disclosure via Symlink Sandbox Bypass github.com/klauspost/compress: klauspost/compress: Denial of Service via integer overflow in dictionary processing com.fasterxml.jackson.core/jackson-core: tools.jackson.core/jackson-core: jackson-core: Denial of Service via excessively long JSON property names accelerate: Hugging Face Accelerate: Path Traversal and Denial of Service via weight_map brace-expansion: DoS via unbounded intermediate arrays, bypassing the CVE-2026-14257 mitigation aiohttp: AIOHTTP: HTTP Request Smuggling via WebSocket Upgrade aiohttp: AIOHTTP: Denial of Ser…&lt;/p&gt;</description>
      <content:encoded>&lt;p&gt;markdown: python-markdown: Denial of Service via inefficient inline link parsing undici: undici: Information disclosure and Denial of Service via malformed Cache-Control directives @fastify/static: @fastify/static: Information disclosure via route guard bypass fast-uri: fast-uri: Host confusion vulnerability via backslash in URI authority undici: undici: Denial of Service via unrequested WebSocket subprotocol extract-zip: extract-zip: Arbitrary file write via symlink in archive encoding/asn1: golang: Go encoding/asn1: Denial of Service via excessive recursion in Unmarshal find-my-way: find-my-way: Denial of Service vulnerability in HTTP/2 server systeminformation: systeminformation: Arbitrary Command Execution via OS Command Injection mlflow: mlflow: Arbitrary file write via FileStore path traversal mcp: mcp: Denial of Service via session memory exhaustion in streamable HTTP servers nltk: NLTK: Information Disclosure via Symlink Sandbox Bypass github.com/klauspost/compress: klauspost/compress: Denial of Service via integer overflow in dictionary processing com.fasterxml.jackson.core/jackson-core: tools.jackson.core/jackson-core: jackson-core: Denial of Service via excessively long JSON property names accelerate: Hugging Face Accelerate: Path Traversal and Denial of Service via weight_map brace-expansion: DoS via unbounded intermediate arrays, bypassing the CVE-2026-14257 mitigation aiohttp: AIOHTTP: HTTP Request Smuggling via WebSocket Upgrade aiohttp: AIOHTTP: Denial of Ser…&lt;/p&gt;</content:encoded>
      <guid isPermaLink="false">https://cve.radiocsirt.org/vuln/rhsa-2026:95872</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>
