{
 "id": "AIR-2025-0287",
 "slug": "cve-2025-46570-timing-side-channel-in-vllm-prefix-caching",
 "title": "CVE-2025-46570: Timing Side-Channel in vLLM Prefix Caching",
 "summary": "vLLM contains a timing side-channel vulnerability linked to its chunk-based prefix caching mechanism. Differences in time to first token allow observers to infer whether specific sensitive prompts were previously cached.",
 "type": "attack_on_ai",
 "categories": [
  "infra_vuln"
 ],
 "severity": 2,
 "targets": [
  "framework_or_tooling"
 ],
 "products": [
  "vLLM"
 ],
 "incident_date": "2025-05-28",
 "incident_date_precision": "day",
 "first_reported": "2025-05-28",
 "sources": [
  {
   "url": "https://github.com/advisories/GHSA-4qjh-9fv9-r85r",
   "publisher": "GitHub Advisory Database",
   "title": "CVE-2025-46570: Potential Timing Side-Channel Vulnerability in vLLM’s Chunk-Based Prefix Caching",
   "published_at": "2025-05-28"
  },
  {
   "url": "https://nvd.nist.gov/vuln/detail/CVE-2025-46570",
   "publisher": "NVD",
   "title": "CVE-2025-46570: vLLM is an inference and serving engine for large language models (LLMs).",
   "published_at": "2025-05-29"
  }
 ],
 "source_count": 2,
 "confidence": 0.95,
 "status": "published",
 "origin": "seed_ghsa",
 "tags": [
  "CVE-2025-46570",
  "side-channel",
  "vllm"
 ],
 "created_at": "2026-10-18T07:00:14.101Z",
 "updated_at": "2026-10-18T10:04:21.885Z"
}