Use of Weak Hash Affecting vllm package, versions [,0.31.0)
Threat Intelligence
The probability is the direct output of the EPSS model, and conveys an overall sense of the threat of exploitation in the wild. The percentile measures the EPSS probability relative to all known EPSS scores. Note: This data is updated daily, relying on the latest available EPSS model version. Check out the EPSS documentation for more details.
Do your applications use this vulnerable package?
In a few clicks we can analyze your entire application and see what components are vulnerable in your application, and suggest you quick fixes.
Test your applications- Snyk IDSNYK-PYTHON-VLLM-21347786
- published21 Oct 2026
- disclosed21 Oct 2026
- creditClinton Thomas
Introduced: 21 Oct 2026
NewCVE-2026-114054 (opens in a new tab)How to fix?
Upgrade vllm to version 0.31.0 or higher.
Overview
vllm is an A high-throughput and memory-efficient inference and serving engine for LLMs
Affected versions of this package are vulnerable to Use of Weak Hash via MultiModalHasher.serialize_item and iter_item_to_bytes in vllm/multimodal/hasher.py, where the hash digest input stream is not framed with per-chunk lengths or container-kind tags. Because distinct byte sequences can serialize to identical concatenated bytes, two different multimodal inputs can produce the same cache key, causing incorrect cache hits. An authenticated attacker can craft multimodal inputs that collide with a victim's cached entry, causing the wrong cached result to be served or evicting valid cache entries, resulting in a crash or incorrect inference output that constitutes a denial of service.