CVE-2026-94627
HIGHDescription
vLLM Mooncake connector through 0.29.0 fails to properly manage GPU KV cache block ownership when concurrent child requests share a single transfer ID in prefill/decode disaggregated deployments. Attackers can trigger GPU memory exhaustion by submitting completion requests with multiple prompts, causing orphaned KV cache blocks to accumulate until process restart and eventually preventing legitimate requests from executing.
Is your site exposed to CVE-2026-94627?
Run a free security scan — no signup, results in seconds.
CVSS v3.1 Score
EPSS — Exploit Prediction
EPSS estimates the probability that this vulnerability will be exploited in the wild within the next 30 days. A higher score means more likely to be exploited.
Weakness Type (CWE)
References
Other References
Frequently Asked Questions
What is CVE-2026-94627? +
How severe is CVE-2026-94627? +
How do I check if I'm vulnerable to CVE-2026-94627? +
Related Vulnerabilities
A denial-of-service security issue exists within the 1794-AENTR adapter due to improper memory handling of CIP protocol requests. This vulnerability …
Virtual attribute handling in Ping Identity PingDirectory in affected versions allows only authorized users to exhaust java memory heap when …
Missing release of memory after effective lifetime, Missing release of resource after effective lifetime vulnerability in Apache Thrift THeaderTransport. This …
Missing Release of Memory after Effective Lifetime vulnerability in leandrocp mdex and mdex_native allows an attacker who controls a rendered …
Missing release of memory after effective lifetime vulnerability in Apache Thrift c++ bindings. This issue affects Apache Thrift: before 0.25.0. …
Missing release of memory after effective lifetime vulnerability in Softing smartLink allows resource leak exposure. This issue affects smartLink HW-PN: …