Live Exposure Defense: From CVE to Confirmed Exposure in 12 Hours – See more

Back
New CVE Detected

CVE-2026-103765 – Unauthenticated Metadata Manipulation Enabling KV Cache Transfer Redirection – Moo

Be the first to know when new zero-days emerge:

Subscribe to Email Alerts
Subscribe to RSS Feed

Summary

CVE-2026-103765 is a missing-authentication vulnerability (CWE-306) in Mooncake’s HTTP metadata server, used by the mooncake-transfer-engine component to coordinate KV cache transfers between distributed LLM inference nodes. The /metadata handler accepts unauthenticated GET, PUT, and DELETE requests, letting any network-reachable attacker read, overwrite, or delete transfer engine metadata — including segment descriptors that control where KV cache payloads are sent. This can be abused to redirect live KV cache transfers (potentially containing sensitive prompt/context data) to an attacker-controlled listener, or to disrupt cluster operation. The issue is rated High/Critical severity (CVSS 8.8) and affects all versions of Mooncake through 0.3.13.post1.

Technical details

  • Root cause: The HTTP metadata server (mooncake/http_metadata_server.py, run via python -m mooncake.http_metadata_server, default bind 0.0.0.0:8080) implements GET/PUT/DELETE handlers for the /metadata endpoint with no authentication, authorization, or input validation. This server acts as the documented backend used by TransferEngine‘s HTTPStoragePlugin (src/transfer_metadata_plugin.cpp) via connection strings such as http://host:28080/metadata.
  • Trigger conditions: Any host that can reach the metadata server’s HTTP port can interact with it — no credentials, tokens, or prior session state are required.
  • Attack vector: Network. An attacker sends a GET request to read a victim engine’s segment descriptor (exposing absolute buffer addresses and TCP data ports), then PUTs a modified descriptor (e.g., poisoning tcp_data_port) to redirect that engine’s outbound KV cache transfers to an attacker-controlled listener. A duplicate-key protection on rpc_meta entries can be bypassed by issuing a DELETE followed by a PUT, allowing re-creation of metadata under attacker control.
  • Impact: Full confidentiality, integrity, and availability compromise of the KV cache transfer path — senders report successful transfers while data actually reaches the attacker’s listener (data exfiltration of prompts/context), cluster metadata can be deleted to break new transfers (denial of service), and unbounded PUT operations can exhaust server memory.

Affected software

  • Vendor/Project: kvcache-ai / Mooncake
  • Package: mooncake-transfer-engine (PyPI)
  • Affected versions: All versions through 0.3.13.post1, including the current main branch example server (mooncake-wheel/mooncake/http_metadata_server.py)

Severity

  • CVSS v4.0: 8.8 (High) — CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:L/VA:H/SC:N/SI:N/SA:N
  • CVSS v3.1: 9.4 (Critical) — CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:L/A:H
  • No authentication (PR:N), no user interaction (UI:N), and network attack vector (AV:N) make this remotely exploitable without any prerequisites.

Mitigation and recommended actions

  • Immediate: As of this writing, the maintainers have not released a patched version — treat any deployment of the HTTP metadata server as unauthenticated-by-design and monitor the upstream kvcache-ai/Mooncake repository (GitHub issue #4444) for a fix.
  • Workarounds / network mitigations:
    • Do not expose the Mooncake HTTP metadata server (default port 8080, or any configured metadata port such as 28080) to untrusted networks or the public internet.
    • Restrict access to the metadata server strictly to trusted hosts within the inference cluster using network segmentation, firewall rules, or a service mesh with mutual TLS.
    • Where supported, replace the unauthenticated HTTP metadata backend with an authenticated alternative (e.g., etcd with access controls) for the TransferEngine‘s metadata plugin.
    • Monitor metadata server logs for unexpected GET/PUT/DELETE requests on /metadata from unrecognized sources, and for anomalous changes to segment descriptors (tcp_data_port values pointing outside the expected cluster range).

References

Are you exposed?

Get a free report of your organization’s exposure to this CVE and threat

  • Mapping of all Assets with this Technology
  • Identification of potentially exposed assets to this CVE
  • Confirmation of verified exploitable assets
Get Exposure Report

How IONIX’s External Exposure Management Platform Detects and Validates
Zero-Days to Shrink MTTR

1

Map your entire attack surface (continously)

IONIX uses multi-factor discovery methods, including DNS analysis, certificate mapping, metadata inspection, and more, to automatically map every internet-facing asset across your environment. This includes cloud instances, third-party platforms, shadow IT, and even forgotten infrastructure that traditional tools miss.

2

Monitor for new CVEs

Dozens of threat intel feeds using agentic technology are continuously analyzed to detect the appearance of proof-of-concept code, exploit kits, and indicators of active targeting. IONIX goes further by applying AI to proactively evaluate whether emerging vulnerabilities are likely to be exploited, even before PoCs go public.

3

Identify Potential External Exposures

Not all CVEs matter. IONIX filters vulnerabilities by asking attacker-centric questions: Can it be reached from the internet? Does it require authentication? Is it being exploited in the wild? This dramatically reduces noise and focuses teams on threats that can actually be weaponized.

4

Create Safe, Scalable Exploit Validations

IONIX transforms real-world PoCs into safe, non-intrusive test payloads that can be run in production environments without disruption. These simulations are precisely targeted to the systems that are vulnerable, ensuring rapid validation without unnecessary load.

5

Execute Exploit Validations

By combining context about software stack, versioning, exposure status, and reachability, IONIX ensures that only the right payloads are executed against the right assets, maximizing efficiency and minimizing risk.

6

Drive Fast and Actionable Remediation

Results are routed through integrations with ticketing, SOAR, and SIEM tools. Issues are written in plain language, bundled into remediation clusters, and prioritized based on asset criticality, exploitability, and blast radius. This shortens mean time to remediation (MTTR) and empowers teams to act with confidence.

Are you exposed?

Get a free report of your organization’s exposure to this CVE and threat

  • Mapping of all Assets with this Technology
  • Identification of potentially exposed assets to this CVE
  • Confirmation of verified exploitable assets
Get Exposure Report
Close

Subscribe to Threat Center RSS

Copy/paste the link below into your preferred RSS reader or follow these instructions to subscribe to Slack alerts.

Open RSS feedSlack setup instructions

Close

Get Real-Time CVE Alerts to Your Email

Be the first to know when new zero-days emerge