CVE-2026-76850

CRITICALPre-NVD 9.89.8—
EchelonGraph scoreHIGH confidence

Score 9.8 from GitHub Security Advisory (severity: CRITICAL) published 2026-08-20. a secondary CVSS source baseline 9.8; sources differ by 0.0.

Triggered by: GitHub Security Advisory CVSS
Sources: epss, ghsa, secondary
Trending — 4 sources updated this weekElevated
9.8EG
EchelonGraph verdictPlan a fixSerious severity, but no confirmed exploitation yet.
  • High severity, but no confirmed exploitation yet
CISA-KEV: Not listedEPSS PROB: 1%CVSS: 9.8Exploit: Elevated riskExposed: 0

No vendor fix yet — apply a workaround or compensating control (WAF / firewall / segmentation) and watch for a patch.

LMDeploy deserializes disaggregated-serving peer messages with pickle. The handle_zmq_recv coroutine in lmdeploy/pytorch/disagg/conn/engine_conn.py reads peer-to-peer cache-free requests with recv_pyobj(), which deserializes the received bytes with pickle.loads(), and the isinstance check against DistServeCacheFreeRequest runs only after deserialization has already completed. The peer that supplies those bytes is caller-controlled: p2p_connect passes remote_engine_endpoint_info.zmq_address from the request body to connect() on the ZMQ PULL socket, and the POST /distserve/p2p_initialize and /distserve/p2p_connect endpoints in lmdeploy/serve/openai/api_server.py apply no authentication unless the server is started with api_keys, which defaults to None. A remote attacker can direct an engine to pull from a ZMQ endpoint under their control and execute arbitrary code in the engine process. Deployments that do not enable disaggregated serving are not affected, because the receive loop is only started once the migration backend accepts the connection.

CVSS v3
9.8
EG Score
9.8(high)
EG Risk
75(Attend)
EG Risk 75/100SSVC: Attend

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 computed
Severity98% × 45%
Exploitation40% × 40%
Automatability100% × 15%
Action: Remediate soon — notable exploitation risk.
EPSS PROB
1%
EPSS %ILE
60%
KEV
Not listed

Published

August 19, 2026

Last Modified

August 20, 2026

Advisory Details (7)

Auto-updated Aug 20, 2026
Patch available. Sources: github_commit, github_release.
generic

[Bug] Unauthenticated pickle deserialization via ZMQ in disaggregated serving → RCE · Issue #4804 · InternLM/lmdeploy · GitHub

https://github.com/InternLM/lmdeploy/issues/4804
generic

LMDeploy Remote Code Execution via Unsafe Pickle Deserialization in the Disaggregated Serving Peer Connector | Advisories | VulnCheck

https://www.vulncheck.com/advisories/lmdeploy-remote-code-execution-via-unsafe-pickle-deserialization-in-the-disaggregated-serving-peer-connector
github_release Patch Available

v0.16.0

Patch available: InternLM/lmdeploy v0.16.0

https://github.com/InternLM/lmdeploy/releases/tag/v0.16.0
github_commit

commit f05b4ad8bf2e (InternLM/lmdeploy)

Fix landed in InternLM/lmdeploy commit f05b4ad8bf2e — awaiting tagged release

https://github.com/InternLM/lmdeploy/commit/f05b4ad8bf2e2d84101a1d63b3c44fadd99223b2
generic

lmdeploy/lmdeploy/pytorch/disagg/conn/engine_conn.py at v0.15.0 · InternLM/lmdeploy · GitHub

https://github.com/InternLM/lmdeploy/blob/v0.15.0/lmdeploy/pytorch/disagg/conn/engine_conn.py#L79
generic

lmdeploy/lmdeploy/pytorch/disagg/conn/engine_conn.py at v0.15.0 · InternLM/lmdeploy · GitHub

https://github.com/InternLM/lmdeploy/blob/v0.15.0/lmdeploy/pytorch/disagg/conn/engine_conn.py#L61
generic

GitHub - InternLM/lmdeploy: LMDeploy is a toolkit for compressing, deploying, and serving LLMs. · GitHub

https://github.com/InternLM/lmdeploy

Vendor Advisories for CVE-2026-76850(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 17× in last 7d / 17× 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.

  1. 2026-08-21 08:43 UTCGHSA enrichment
  2. 2026-08-21 04:24 UTCGHSA enrichment
  3. 2026-08-21 00:06 UTCEG score recompute
  4. 2026-08-21 00:05 UTCGHSA enrichment
  5. 2026-08-20 22:56 UTCEPSS rescore
  6. 2026-08-20 19:46 UTCGHSA enrichment
  7. 2026-08-20 15:28 UTCEG score recompute
  8. 2026-08-20 15:28 UTCGHSA enrichment
  9. 2026-08-20 14:27 UTCEG score recompute
  10. 2026-08-20 14:27 UTCGHSA enrichment
  11. 2026-08-20 11:17 UTCGHSA enrichment
  12. 2026-08-20 06:59 UTCGHSA enrichment
  13. 2026-08-20 02:40 UTCEG score recompute
  14. 2026-08-20 02:40 UTCGHSA enrichment
  15. 2026-08-19 22:20 UTCEG score recompute
  16. 2026-08-19 21:48 UTCEG score recompute
  17. 2026-08-19 21:47 UTCMITRE cvelistV5first tracked

Related CVEs(same CWE)

Frequently asked(5)

What is CVE-2026-76850?
CVE-2026-76850 is a critical vulnerability published on August 19, 2026. LMDeploy deserializes disaggregated-serving peer messages with pickle. The handlezmqrecv coroutine in lmdeploy/pytorch/disagg/conn/engineconn.py reads peer-to-peer cache-free requests with recvpyobj(), which deserializes the received bytes with pickle.loads(), and the isinstance check against…
When was CVE-2026-76850 disclosed?
CVE-2026-76850 was first published in the National Vulnerability Database on August 19, 2026, with the most recent update on August 20, 2026. EchelonGraph re-ingests CVE updates from NVD on a 2-hour cycle, so this page reflects the latest published state.
Is CVE-2026-76850 actively exploited?
CVE-2026-76850 is not currently on CISA's Known Exploited Vulnerabilities catalog. FIRST EPSS estimates a 1% probability of exploitation in the next 30 days, which ranks it in the top 40.4% of all scored CVEs.
What is the CVSS score of CVE-2026-76850?
CVE-2026-76850 has a CVSS v3 base score of 9.8 (NVD).
How do I remediate CVE-2026-76850?
Patch to the fixed version published by the affected vendor. Where vendor advisories exist for CVE-2026-76850, EchelonGraph cross-links them in the Vendor Advisories panel below — those typically contain the canonical remediation steps, fixed version numbers, and any vendor-specific mitigations.

Dependency Blast Radius

Explore the affected products and dependency analysis for CVE-2026-76850

Explore →

Is Your Infrastructure Affected by CVE-2026-76850?

EchelonGraph automatically scans your cloud infrastructure and maps CVE exposure using blast radius analysis.

Start Free Scan →Browse All CVEs