vLLM is an inference and serving engine for large language models. Prior to 0.28.0, the default mirrored multimodal LRU cache can commit a media hash in the frontend sender cache during multimodal rendering and before engine admission, while the engine receiver cache never receives the payload if that request is rejected. A later request reusing the same media hash causes MultiModalProcessorSenderCache to send no payload and MultiModalReceiverCache to reach an assertion with the message "Expected a cached item," producing a shared-service availability failure. This issue is fixed in version 0.28.0.
CVE-2026-105753
MEDIUMPre-NVD 6.56.5
EchelonGraph scoreLOW confidence
This medium-severity CVE scores 6.5 under a secondary CVSS source (NVD's own analysis pending). EPSS exploit-prediction score not yet available (the EPSS model rescores nightly; freshly-published CVEs typically appear within 48 hours). GitHub Security Advisory data not yet ingested — confidence will rise once GHSA publishes (typical lag: hours to days for open-source ecosystem CVEs; never for infrastructure-only CVEs).
Triggered by: NVD CVSS baseline
Sources: secondary
6.5EG
EchelonGraph verdictMonitorLow exploitation likelihood right now — keep watching.
- Lower severity and no public exploit yet
CISA-KEV: Not listedEPSS PROB: —CVSS: 6.5Exploit: None knownExposed services: Not assessed
A fix is available — apply it.
- CVSS v3
- 6.5
- EG Score
- 6.5MEDIUMlow confidence
- EG Risk
- 34EG Risk 34/100CISA SSVC
EG Risk is EchelonGraph's 0–100 priority score: it fuses intrinsic severity with real-world exploitation and automatability so you can rank equal-severity CVEs and fix the most dangerous first. Higher = act sooner. Distinct from the 0–10 EG Score (severity).
How it’s computedSeverity65% × 45%Exploitation0% × 40%Automatability30% × 15%CISA SSVC: Track at low or medium mission impact; Track or Attend at high (mission-essential systems).Action: A fix is available. Apply it within your standard update timelines at low or medium mission impact and sooner than that at high. - EPSS PROB
- —
- EPSS %ILE
- —
- KEV
- Not listed
CISA SSVCTrack at low or medium mission impact; Track or Attend at high (mission-essential systems).
A fix is available. Apply it within your standard update timelines at low or medium mission impact and sooner than that at high.
Exploitation none (no KEV listing, exploit record or EPSS ≥ 50%) · Automatable unknown (not published for this CVE) · Technical impact partial (CVSS below 9.0). Mission impact is CISA's Mission & Well-being decision point, and only you can judge it: high means the affected system is essential to your organisation's mission, or its compromise could cause irreversible harm to people. CISA's decision table
Published
October 5, 2026
Last Modified
October 5, 2026
Advisory Details (5)
Auto-updated Oct 5, 2026Patch available. Sources: github, github_pr, github_commit, github_release.
github_release Patch Available
v0.28.0
Patch available: vllm-project/vllm v0.28.0
https://github.com/vllm-project/vllm/releases/tag/v0.28.0github_commit
commit 396204230423 (vllm-project/vllm)
Fix landed in vllm-project/vllm commit 396204230423 — awaiting tagged release
https://github.com/vllm-project/vllm/commit/396204230423b7cc6798300926b8fa30190d26a9github_pr
Roll back mirrored multimodal cache entries after rejection
Upstream fix in progress (PR #51897 · vllm-project/vllm, state=closed)
https://github.com/vllm-project/vllm/pull/51897github_pr
[Bugfix][V1][Multimodal] Recover from P0/P1 processor cache drift (#46747)
Fix merged in vllm-project/vllm PR #46747 on 2026-08-12 — awaiting tagged release
https://github.com/vllm-project/vllm/pull/46747githubMentions a patch (unverified)
Mirrored multimodal IPC caches desync after a rejected request — a later request reusing the same media hash trips a receiver assertion in the engine core · Advisory · vllm-project/vllm · GitHub
https://github.com/vllm-project/vllm/security/advisories/GHSA-ph3r-5jfg-f84fVendor Advisories for CVE-2026-105753(1)
These vendors published their own advisory mentioning this CVE — often with vendor-specific remediation steps + affected product lists not in NVD.
Weakness Classification(1)
MITRE Common Weakness Enumeration — the root-cause categories this CVE belongs to.
Data Freshness Timeline
(refreshed 3× in last 7d / 3× in last 30d)
Each row is a source pipeline that fetched or updated this CVE on that date, with what changed. For example, "NVD update" means NVD published or revised its analysis for this CVE; "MITRE cvelistV5" means we ingested or refreshed it from the CNA feed. Most recent first.
- 2026-10-05 23:27 UTCEG score recompute
- 2026-10-05 22:46 UTCEG score recompute
- 2026-10-05 22:45 UTCMITRE cvelistV5first tracked
Related CVEs(same CWE)
Frequently asked(4)
What is CVE-2026-105753?
CVE-2026-105753 is a medium vulnerability published on October 5, 2026. vLLM is an inference and serving engine for large language models. Prior to 0.28.0, the default mirrored multimodal LRU cache can commit a media hash in the frontend sender cache during multimodal rendering and before engine admission, while the engine receiver cache never receives the payload if…
When was CVE-2026-105753 disclosed?
CVE-2026-105753 was first published on October 5, 2026. EchelonGraph re-ingests CVE updates from NVD on a 2-hour cycle, so this page reflects the latest published state.
What is the CVSS score of CVE-2026-105753?
CVE-2026-105753 has a CVSS base score of 6.5 (a secondary CVSS source that NVD displays; NVD's own analysis pending). The EG score is currently aggregating — additional source signals are being incorporated as they become available..
How do I remediate CVE-2026-105753?
A fix for CVE-2026-105753 is available: update to the fixed version the vendor names in its advisory. The vendor advisories EchelonGraph has for CVE-2026-105753 are linked in the Vendor Advisories panel on this page.
Dependency Blast Radius
Explore the affected products and dependency analysis for CVE-2026-105753
Is Your Infrastructure Affected by CVE-2026-105753?
EchelonGraph automatically scans your cloud infrastructure and maps CVE exposure using blast radius analysis.