Black Friday60% off BLACKFRIDAY2026
PlatformPlansPartners
Resources
ToolsBlogDocs
Sign Up →
critical·Product Based Web Vulnerabilities·Updated Nov 17, 2026

CVE-2026-102895 Scanner

CVE-2026-102895 Scanner - Unauthorized Admin Access vulnerability in GPUStack

Est. Time~10 seconds
Scan TypeGroup Scan
Targetsdomain, subdomain, ipv4
CostFree
3
Times Used
continuous scan runs
4.2k
Continuously Checked
assets under CS
0
Vulnerabilities Found
confirmed findings
References
🔴
CISA Known Exploited Vulnerability
This CVE is actively exploited in the wild. CISA mandates federal agencies to patch immediately.
9.8
CVSScritical
Exploitable remotely over the internet · no authentication required.
Description

GPUStack is an open-source GPU cluster manager for AI model serving and GPU instance provisioning. Prior to 2.2.0, get_current_user in gpustack/api/auth.py treated a loopback request source as the built-in administrator when force_auth_localhost was false, which was the default. A remote request could appear to originate from 127.0.0.1 because client_ip_getter accepted a same-host reverse proxy peer or a client-supplied X-GPUStack-Real-IP loopback value. The bypass allowed unauthenticated administrative access, including user and model changes, persistent administrator API-key creation, disclosure of worker and cloud-provider secrets, and code execution through POST /v2/inference-backends/from-yaml. Exploitation was observed in the wild against a 2.1.1 deployment. This issue is fixed in version 2.2.0.

Attack Vector
Network
Privileges Req.
None
User Interaction
None
Affected
gpustackby gpustack
< 2.2.0
Updated Nov 17, 2026View on NVD →
Detail

GPUStack is a cloud-native platform used primarily for high-performance computing tasks, designed to manage clusters of GPUs efficiently for parallel processing workloads. It is often employed by research institutions and companies needing substantial computational resources to perform machine learning, AI development, and scientific computations. Administrators typically utilize GPUStack to streamline and automate the deployment, scaling, and management of GPU resources across numerous nodes. Its interface is designed to cater to both technical and non-technical users, providing tools for configuration, monitoring, and diagnostics. GPUStack facilitates integration with various existing systems and workflows, allowing for more efficient utilization of computation resources. Organizations rely on the platform to maximize their investment in GPU technology while minimizing the complexity of managing such infrastructure.

The vulnerability in question involves the potential for an attacker to gain unauthorized administrative access to GPUStack versions before 2.2.0 by exploiting improperly trusted HTTP headers. Specifically, the issue arises because the software may misconstrue hostile X-GPUStack-Real-IP headers as a legitimate local request, granting admin access unwarrantedly. Malicious users can leverage this exploit to bypass authentication mechanisms, compromising system integrity. The oversight essentially turns any incoming request into one that appears to originate from a trusted source unless additional protections are in place. This flaw undermines the boundaries between trusted internal communications and untrustworthy external requests. Efforts must be made by users of affected systems to fortify their defense mechanisms until a proper upgrade can be implemented.

From a technical standpoint, the affected parameter is the X-GPUStack-Real-IP header, which should typically be restricted to communications verified as originating from the same local network. However, GPUStack software up to version 2.2.0 improperly processes this information, potentially granting outsiders admin-level permissions without due verification. This vulnerability is exacerbated when requests are forwarded by reverse proxy setups that inadvertently signal GPUStack to trust them as local. Until version 2.2.0, anyone able to manipulate packet headers to include the poorly vetted X-GPUStack-Real-IP information could take advantage of this oversight. It is crucial that configurations be updated to ensure all incoming data packets receive appropriate scrutiny, particularly regarding their origins. Recommended practices involve systematic updates to routing and filtering rules, particularly where these interfaces intersect with public or less secure networks.

When malicious actors exploit this vulnerability, the consequences can be severe. Unauthorized parties could gain administrative privileges, offering them the capacity to alter or commandeer resource management configurations and workflows. The credibility and security of sensitive computational processes are jeopardized, potentially leading to data breaches or the unauthorized execution of harmful code. Further, rogue modifications to configuration files could destabilize system operations, causing disruptions or undesirable behavior across the networked resources. This situation could manifest in unexpected downtimes or compromised data integrity, undermining trust in the infrastructure's reliability. Preventative measures, therefore, are essential in pre-emptively safeguarding organizational resources from these risks.

REFERENCES

Solution Advice
  • Upgrade to GPUStack 2.2.0 or later to secure against this vulnerability.
  • Enable force_auth_localhost to prevent unauthorized administrative access.
  • Implement strict rules to strip untrusted X-GPUStack-Real-IP headers at the network edge.
  • Restrict API access through robust firewall configurations.
  • Regularly audit and monitor access logs for unusual activity indicative of an exploit attempt.

Get AI-powered remediation steps tailored to your asset.

Try AI Solutions →

Check your infrastructure.
Right now.

12,000+ scanners. Free to start. No credit card required.