CVE-2026-56340
Published: 20 June 2026
Summary
CVE-2026-56340 is a high-severity Improper Input Validation (CWE-20) vulnerability in Vllm Vllm. Its CVSS base score is 8.7 (High).
Operationally, exploitation aligns with the MITRE ATT&CK technique Exploit Public-Facing Application (T1190); ranked at the 27.2th percentile by exploit likelihood (below the median); it is not currently listed in the CISA KEV catalog; a public proof-of-concept is referenced.
OWASP Top 10 for Web (2025)
EU & UK References
- 🇪🇺 ENISA EUVD: EUVD-2026-38129
Vulnerability details
vLLM versions >= 0.10.2 and < 0.13.0 are missing sparse tensor validation in multimodal embeddings processing. Because PyTorch disables sparse tensor invariant checks by default, an attacker can submit crafted embedding requests with malformed (negative or out-of-bounds) tensor indices, when…
more
the prompt-embeds feature is enabled, to trigger crashes or resource exhaustion (denial of service), with potential for out-of-bounds/write-what-where memory corruption. This continues CVE-2025-62164, whose prior fix only disabled the feature by default rather than addressing the root cause.
- CWE(s)
Related Threats
MITRE ATT&CK Enterprise TechniquesAI
Why these techniques?
CWE-20 input validation flaw in public vLLM inference API directly enables remote exploitation (T1190) via crafted multimodal embedding requests, resulting in application crashes or resource exhaustion (T1499.004).
Affected Assets
Mitigating Controls
Likely Mitigating Controls AI
Per-CVE control mapping for this CVE has not run yet; the list below is derived from the weakness types (CWEs) cited in the NVD entry.
Security testing and developer training directly verify and enforce proper input validation, reducing exploitability of injection and malformed-data weaknesses.
Security testing and evaluation at multiple SDLC stages directly detects missing or flawed input validation, with the required remediation process ensuring fixes are applied.
Directly implements checks on information inputs to reject invalid data before processing.
Out-of-bounds writes that corrupt control flow or inject shellcode are rendered non-executable by the same memory protections.
Spam protection mechanisms perform filtering and detection on inbound/outbound messages, directly compensating for missing or weak input validation of unsolicited content.