Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessObtain low-privilege UFM credentials
-
DeliveryReach UFM management API from adjacent network
-
ExploitCraft malicious plugin API request with injected commands
-
ExecutionUFM passes unsanitized input to system shell
-
PersistExecute arbitrary code on UFM host
-
ImpactEscalate privileges and exfiltrate fabric data
Vulnerability Assessment
| Weakness |
CWE
OWASP Top 10:2025
A05:2025 - Injection
ASVS 5.0
V1 - Encoding and Sanitization Command injection prevention
MITRE ATT&CK
1 technique across 1 tactic: Execution The technique list with ATT&CK links and the kill-chain view is part of the Basic plan and up. |
| Exploitation Conditions | The plugin management API endpoint must be network-reachable from the attacker's position, requiring adjacency to the UFM management network segment (AV:A - same VLAN, broadcast domain, or directly connected management interface). The attacker must possess valid UFM Enterprise credentials at a low-privilege level (PR:L); unauthenticated access is not sufficient. No user interaction on the server side is required (UI:N), and attack complexity is low (AC:L), meaning no race conditions, timing dependencies, or special environmental states are needed beyond the above. Exploitation is bounded by: the adjacent-network requirement (not reachable from the internet in a properly firewalled deployment), the need for at least low-privilege authentication to the UFM API, and UFM's niche deployment in InfiniBand-connected HPC and AI environments. Organizations that restrict the plugin management API to administrator accounts only and enforce management VLAN isolation eliminate or significantly reduce the practical attack surface. |
| Impact | The CVSS 8.0 score reflects high impact across all three pillars (C:H/I:H/A:H), but two factors meaningfully constrain real-world exploitability: the Adjacent attack vector (AV:A) and the low-privilege authentication requirement (PR:L). AV:A limits exposure to actors who are already on the same network segment or management VLAN as the UFM appliance - internet-based attackers cannot reach this directly in a properly segmented environment. PR:L requires the attacker to first obtain valid credentials, even at a low permission level; unauthenticated exploitation is not possible per the CVSS vector. EPSS data is absent from the provided intelligence, and no CISA KEV listing or public POC has been identified, so exploitation appears theoretical at present. The realistic threat actor is an insider, a compromised service account, or an attacker who has already penetrated the management plane. Given UFM's role in orchestrating InfiniBand fabric - which underpins GPU cluster interconnects in AI/HPC environments - a successful compromise could cascade into cluster-wide availability disruption or credential exfiltration. The adjacent-network restriction is load-bearing: organizations that enforce strict management-plane segmentation and least-privilege API access significantly reduce realistic risk despite the high CVSS base score. |
| Exploit ScenarioDerived | An attacker who has obtained low-privileged UFM Enterprise credentials - through phishing, credential stuffing against a management host, or insider access - connects from within the management VLAN and issues a crafted HTTP API request to the plugin management endpoint, embedding shell metacharacters or command separators in a plugin name or configuration parameter. The UFM application passes the unsanitized input to a system command without escaping, resulting in execution of attacker-supplied commands under the UFM process context. The attacker then leverages this execution to escalate to root, exfiltrate fabric topology maps and stored credentials, or corrupt fabric configuration - disrupting the InfiniBand interconnect for dependent GPU compute nodes. No public POC has been identified, but CWE-77 injection patterns are well-understood and the API surface is documented. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Aug 25, 2026
Code injection in NVIDIA Unified Fabric Manager Enterprise's plugin management API allows low-privileged, adjacent-network authenticated users to execute arbitrary commands on the UFM host by submitting a specially crafted API request. All current release tracks - GA, LTS 2023, LTS 2024, and LTS 2025 - are confirmed affected per vendor-supplied CPE data. Successful exploitation yields code execution, privilege escalation, and information disclosure on a platform that controls InfiniBand fabric infrastructure, amplifying downstream impact on managed HPC and AI compute clusters. No public exploit code or CISA KEV listing has been identified at time of analysis.
RemediationAI
The primary remediation is to apply the vendor-released patch for the applicable UFM Enterprise release track (GA, LTS 2023, LTS 2024, or LTS 2025). No specific fixed version numbers are derivable from the available input data; consult the NVIDIA Product Security advisory at https://github.com/NVIDIA/product-security/tree/main/2026/5809 for exact patched build identifiers per release track. Pending patch deployment, enforce network segmentation so the UFM management API is reachable only from dedicated, hardened management hosts - this directly mitigates the AV:A attack vector by preventing unauthorized adjacent-network access. Additionally, restrict plugin management API access to administrator-level accounts only, removing low-privileged user access to this endpoint; note that this may limit operational workflows for teams relying on delegated plugin management. Audit all existing low-privilege UFM accounts for necessity and rotate credentials as a precautionary measure given the authentication-bypass-to-injection risk. Implement API request logging on the plugin management endpoint to detect anomalous or crafted inputs.
Affected VersionsENISA
Affected Packages
Source: OSV (Open Source Vulnerabilities)
Affects your stack
Affects 4 packages in euvd
Checking this CVE against your own stacks and servers is part of the Basic plan and up.
Technical ContextAI · show details
NVIDIA UFM (Unified Fabric Manager) Enterprise is a specialized network fabric management platform for InfiniBand-based HPC and AI data center interconnects, responsible for topology discovery, telemetry, and fabric provisioning. The vulnerability is rooted in CWE-77 (Improper Neutralization of Special Elements used in a Command - Command Injection), meaning user-controlled input reaching the plugin management API is incorporated into a downstream system command without sufficient sanitization or escaping. The CVSS vector's AV:A (Adjacent) metric is consistent with UFM management interfaces typically being reachable only from within the management VLAN or InfiniBand fabric segment, not from the open internet. All four CPE-confirmed release tracks share the flaw: cpe:2.3:a:nvidia:unified_fabric_manager_enterprise_-_ga, cpe:2.3:a:nvidia:unified_fabric_manager_enterprise_-_lts_2025, cpe:2.3:a:nvidia:unified_fabric_manager_enterprise_-_lts_2024, and cpe:2.3:a:nvidia:unified_fabric_manager_enterprise_-_lts_2023. The plugin management subsystem is the specific attack surface, suggesting the injection occurs during plugin installation, update, or configuration operations where filenames, identifiers, or parameters are passed to a shell command.
Know when Unified Fabric Manager Enterprise Ga 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
AV:A/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H
Adjacent vector reflects management-VLAN-only reachability; PR:L required per description; full C/I/A impact from command injection with no scope change beyond UFM host.
Primary rating from Vendor (nvidia).
CVSS VectorVendor: nvidia
DescriptionCVE.org
NVIDIA UFM Enterprise contains a vulnerability in the plugin management API, where an authenticated user with low privileges could inject code by sending a specially crafted API request. A successful exploit of this vulnerability might lead to code execution, escalation of privileges and information disclosure.
Articles & Coverage 1
Executive ViewAI
NVIDIA Unified Fabric Manager Enterprise contains a code injection vulnerability (CVE-2026-24169) affecting all release tracks that allows authenticated network users to execute arbitrary commands on systems controlling InfiniBand fabric infrastructure for HPC and AI compute clusters. Organizations using UFM Enterprise face immediate risk to fabric availability, compute cluster integrity, and data confidentiality across all connected systems, with no vendor patch currently available.
No vendor-released patch identified at time of analysis. Interim mitigations: restrict network access to UFM plugin management API through firewall rules or network segmentation; disable or isolate non-critical UFM plugin functionality; implement strict authentication and authorization controls on API endpoints; deploy network-based intrusion detection targeting UFM API endpoints for exploit indicators; maintain detailed audit logging of all API access and administrative actions on UFM systems; and establish incident response procedures for rapid isolation of affected UFM instances if compromise is detected.
Threat Intelligence Overview
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.
What every action level means
SSVC AssessmentENISA
Priority Score
Attack Technique
OS command injection occurs when an application passes unsanitized user input directly into system shell commands.
Information disclosure occurs when an application unintentionally exposes sensitive data that aids attackers in reconnaissance or directly compromises security.
Remote Code Execution represents the critical moment when an attacker successfully runs arbitrary code on a target system without physical access.
Authentication bypass in NVIDIA UFM Enterprise's web interface authorization component allows an attacker on the adjacen
Command injection in NVIDIA UFM Enterprise's IBDiagnet API allows an authenticated administrator with adjacency-level ne
Command injection in NVIDIA UFM Enterprise's user management component enables an authenticated administrator to execute
NVIDIA Unified Fabric Manager Enterprise session management embeds a hard-coded cryptographic key (CWE-321), enabling a
EUVD-2026-65609
GHSA-fp58-r6rm-j2hh
NVD
CVE.org
EPSS
Vulnrichment
Exploit-DB