Skip to main content

Mooncake EUVDEUVD-2026-85753

| CVE-2026-96764 LOW
Allocation of Resources Without Limits or Throttling (CWE-770)
2026-09-24 VulDB GHSA-gqr3-xc87-48w2
2.1
CVSS 4.0 · Vendor: VulDB
View

Attack ChainAIDerived

Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain

  1. Access
    Authenticate to mooncake master RPC
  2. Delivery
    Craft malicious regex input
  3. Exploit
    Call GetReplicaListByRegex
  4. Execution
    Trigger uncontrolled resource allocation
  5. Impact
    Degrade master service availability

Vulnerability Assessment

Weakness
CWE
OWASP Top 10:2025
A06:2025 - Insecure Design
ASVS 5.0
V2 - Validation and Business Logic Resource allocation limits
MITRE ATT&CK

1 technique across 1 tactic: Impact

The technique list with ATT&CK links and the kill-chain view is part of the Basic plan and up.

Exploitation Conditions

Requires network access to the mooncake MasterService RPC endpoint and low-privilege authentication/client access (CVSS PR:L) sufficient to invoke GetReplicaListByRegex. The attacker must supply a manipulated regular expression / query to the Regular Expression Handler that induces uncontrolled resource allocation (CWE-770). Limiting factors: impact is availability-only and rated partial (VA:L, no full crash claimed), the affected product (kvcache-ai mooncake up to 0.3.12/0.3.14-rc1) has a small deployment footprint, exploitation is gated behind low-privilege access rather than being fully unauthenticated, and no vendor patch exists so mitigation depends on restricting RPC exposure and input validation.

Impact

This is a genuine but modest-severity issue. The CVSS 4.0 base of 5.3 (AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:L) reflects a low-privilege, remotely reachable availability-only weakness - no confidentiality or integrity impact. The E:P (Proof-of-Concept) exploit maturity is corroborated by a publicly available exploit gist, so publicly available exploit code exists, but this is NOT the same as active exploitation: there is no CISA KEV listing and no EPSS score provided, so widespread automated exploitation cannot be inferred. CWE-770 (uncontrolled resource allocation) via a regex-processing function (MasterService::GetReplicaListByRegex) is a classic DoS/ReDoS-style pattern where a crafted pattern or matching workload triggers excessive resource consumption; the impact is limited to VA:L (partial availability degradation), not a full node crash per the vendor scoring. The vector states PR:L, so exploitation requires some authenticated/low-privilege access to the master service RPC surface - this meaningfully limits real-world risk versus a fully unauthenticated DoS. Net: a real, verifiable priority for anyone actually running mooncake with exposed master RPC, but low absolute urgency given availability-only impact, low deployment footprint, and no evidence of exploitation beyond a POC.

Exploit ScenarioDerived

A low-privileged client with access to the mooncake master service RPC surface calls GetReplicaListByRegex with a manipulated regular expression or query designed to trigger uncontrolled resource allocation during regex handling. This drives excessive CPU/memory consumption on the master, degrading availability of the KVCache serving layer for other tenants. A public POC exploit is available (VulDB-linked gist), lowering the barrier to reproduction.

Monitor Track and reassess. Why: LOW severity (CVSS 0.1-3.9).
No patch listed yet 0.3% exploitation probability (EPSS)

Get an e-mail when this CVE is exploited, patched or re-scored. Free account.

Alert me

AnalysisAIv1 – Sep 24, 2026

Uncontrolled resource allocation in the MasterService::GetReplicaListByRegex handler of kvcache-ai Mooncake (all versions up to and including 0.3.12, and 0.3.14-rc1) allows a remote low-privilege client that can reach the MasterService RPC endpoint to exhaust server memory or CPU by submitting a crafted regular expression to the Regular Expression Handler. The impact is availability-only and partial (CVSS 4.0 5.3, VA:L; assessed CVSS 3.1 5.3 AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:L), so this is a denial-of-service concern rather than a confidentiality or integrity compromise. Publicly available exploit code exists (a proof-of-concept gist is referenced via VulDB), and the affected project has a small deployment footprint; no CISA KEV entry exists and no vendor patch was released because the vendor did not respond to the disclosure.

RemediationAI

No vendor-released patch identified at time of analysis: the vendor (kvcache-ai) did not respond to the disclosure, and neither VulDB entry 409019 (https://vuldb.com/vuln/409019) nor the CVE record references a fixed release, so there is no exact patched version to cite and operators should not assume a silent fix exists in 0.3.13/0.3.14. Until an upstream fix lands, mitigate by treating the MasterService RPC endpoint as a trusted-only surface: bind it to a private interface or overlay network and firewall it so only known inference clients can reach it, accepting that this breaks legitimate multi-tenant or cross-cluster deployments that rely on remote replica lookups. Add request-side validation and throttling in front of the method - cap regex pattern length, reject or rewrite nested-quantifier/backtracking-prone patterns, cap the number of matched replicas returned, and enforce per-client request timeouts and concurrency quotas - with the trade-off that aggressive regex filtering can break legitimate pattern-based lookups and tight result caps may truncate valid replica lists. Backstop these with OS/container-level resource limits (cgroup memory ceilings, CPU shares, and MasterService process restart supervision) so an abusive query degrades one process rather than the whole inference cluster, and monitor GetReplicaListByRegex latency, memory growth, and result-set size for evidence of exploitation attempts, noting that monitoring is detection rather than prevention. If availability risk is unacceptable for an exposed deployment, the only fully effective option is to disable or remove the regex-based lookup path and use exact-match replica queries instead, at the cost of losing pattern-based replica discovery.

Affected VersionsENISA

mooncake 0.3.14-rc1
mooncake 0.3.2
mooncake 0.3.11
mooncake 0.3.0
mooncake 0.3.5
mooncake 0.3.3
mooncake 0.3.12
mooncake 0.3.6
mooncake 0.3.4
mooncake 0.3.8
mooncake 0.3.9
mooncake 0.3.1
mooncake 0.3.7
mooncake 0.3.10

Affected Packages

euvd mooncake no fix

Source: OSV (Open Source Vulnerabilities)

Affects your stack

Affects 1 package in euvd

Checking this CVE against your own stacks and servers is part of the Basic plan and up.

Technical ContextAI · show details

Mooncake is a KVCache-centric disaggregated serving architecture for LLM inference, in which a MasterService coordinates storage/replica metadata and exposes RPC methods used by clients to locate cached KV blocks across distributed nodes. The vulnerable method, MasterService::GetReplicaListByRegex, accepts a caller-supplied regular expression/query used to match replica entries; the handler does not bound the cost of compiling or executing that pattern, which is the classic CWE-770 (Allocation of Resources Without Limits or Throttling) weakness - regex engines can be driven into exponential backtracking or into materializing very large result sets, and the handler performs no input length, complexity, or result-size throttling. The affected component is therefore the Regular Expression Handler inside the MasterService RPC surface, tracked generically by the CPE cpe:2.3:a:kvcache-ai:mooncake:*:*:*:*:*:*:*:* (all released versions through 0.3.12 and the 0.3.14-rc1 pre-release). Because the CVSS 4.0 vector carries PR:L and the assessed 3.1 vector carries PR:L, exploitation requires an authenticated/low-privilege client session capable of invoking the RPC method over the network (AV:N) - it is not an unauthenticated attack, and no privilege escalation or data disclosure is claimed; the vector is E:P, matching the public proof-of-concept.

Know when Mooncake gets its next vulnerability

  • Watch products and get an email when something new is published
  • Check this CVE against your own stack and servers
  • Exploit and PoC references, threat intelligence from 22 sources

No credit card · 7-day full trial

Severity by source

Vendor (VulDB) PRIMARY 2.1 MEDIUM
CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:L/SC:N/SI:N/SA:N/E:P
vuln.today AI 4.3 MEDIUM

Network-reachable RPC with low-privilege access (PR:L) and no user interaction; regex-driven resource allocation yields partial availability impact only (A:L), no C/I impact.

3.1 AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:L
4.0 AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:L/SC:N/SI:N/SA:N

Primary rating from Vendor (VulDB).

CVSS VectorVendor: VulDB

Attack Vector
Network
Attack Complexity
Low
Attack Requirements
None
Privileges Required
Low
User Interaction
None
Vulnerable system: Confidentiality
None
Vulnerable system: Integrity
None
Vulnerable system: Availability
Low
Subsequent system: Confidentiality
None
Subsequent system: Integrity
None
Subsequent system: Availability
None

Lifecycle Timeline

3
Severity Changed
Sep 24, 2026 - 01:22 NVD
MEDIUM LOW
CVSS changed
Sep 24, 2026 - 01:22 NVD
5.3 (MEDIUM) 2.1 (LOW)
CVE Published
Sep 24, 2026 - 00:15 cve.org
MEDIUM 5.3

DescriptionCVE.org

A weakness has been identified in kvcache-ai mooncake up to 0.3.12/0.3.14-rc1. Impacted is the function MasterService::GetReplicaListByRegex of the component Regular Expression Handler. Executing a manipulation can lead to allocation of resources. The attack may be performed from remote. The exploit has been made available to the public and could be used for attacks. The vendor was contacted early about this disclosure but did not respond in any way.

Threat Intelligence Overview

EPSS Score
0.3%
Top 83% exploitable
Prevalence
Low
niche open-source KVCache/LLM-serving engine, limited production footprint
Exploit
POC
Public exploit code exists
Social Media
1
Sightings
Data Quality
6 sources
Confidence: High
EUVD NVD CVE.org EPSS Vulnrichment CIRCL GitHub VulDB

How this decision was reached

The engine reads each signal below and applies a fixed list of rules from top to bottom. The first rule that matches sets the recommended action – like a statement, every line is shown.

SignalValue
Severity LOW
CISA KEV No
Public POC Yes
EPSS 0.3%
Patch available No
Active threat No
Deciding rule
LOW severity (CVSS 0.1-3.9)
What every action level means
Emergency Immediate action required. Criteria: CRITICAL + in CISA KEV (actively exploited) without patch, OR CRITICAL + EPSS > 30% without patch, OR CRITICAL + active threat signals without patch.
Act Now Prioritize within 24-48h. Criteria: in KEV with patch available, OR CRITICAL + public POC, OR HIGH + EPSS > 10%, OR CRITICAL + active threat, OR any CRITICAL severity.
This Week Schedule remediation this week. Criteria: HIGH severity (CVSS 7.0-8.9). Patch within normal change window.
This Month Plan within 30 days. Criteria: MEDIUM severity (CVSS 4.0-6.9), lower exploitation probability. Include in regular patch cycle.
Monitor Track and reassess. Criteria: LOW severity (CVSS 0.1-3.9), minimal exploitation risk. No immediate action needed.

SSVC AssessmentENISA

SSVC Decision· derived
Scheduled
Exploitation
Poc
System Exposure· derived
Open
Automatable
No
Human Impact· derived
Low
Technical Impact
Partial

Priority Score

30
Low Medium High Critical
KEV: 0
EPSS: 0
CVSS: +10
PoC: +20
Weaponized: 0
Threat intel: 0
Gov advisory: 0

Attack Technique

Denial of Service attacks render applications or systems unavailable by overwhelming resources or triggering failure conditions.

Related Vulnerabilities in kvcache-ai mooncake Access Control and DoS Flaws

View all 3

This site uses cookies essential for authentication and security. No tracking or analytics cookies are used. Privacy Policy