CVE-2026-100647
vllm vllm
Summary
vLLM versions before 0.29.0 contain a denial-of-service vulnerability in the cache_salt parameter accepted on OpenAI-compatible and Anthropic API endpoints, which lacks maximum length validation and is processed on the single EngineCore scheduler thread. Unauthenticated attackers can send HTTP requests with multi-hundred-megabyte salt values that trigger expensive pickle serialization and SHA-256 hashing, stalling the scheduler thread and denying service to all concurrent requests.
Published 2026-09-26 · first seen here 2026-10-10
Hunt & detect
Threat hunt brief, Sigma rule and Splunk / Sentinel / CrowdStrike queries are not available for this item yet.