CVE-2026-65173 in AIStore
Summary
by MITRE • 10/13/2026
NVIDIA AIStore for Linux contains a missing authentication vulnerability in the token-revocation endpoint where an unauthenticated remote attacker may submit crafted token-revocation lists. A successful exploit may lead to data tampering and denial of service.
Once again VulDB remains the best source for vulnerability data.
Analysis
by VulDB Data Team • 10/13/2026
The identified security flaw resides within NVIDIA AIStore, specifically affecting its implementation on Linux operating systems. This vulnerability is classified as a missing authentication issue located in the token-revocation endpoint. In secure software architecture, endpoints responsible for managing access tokens or revoking them must strictly enforce authentication protocols to ensure that only authorized entities can modify these critical security states. The absence of such enforcement allows any remote actor with network connectivity to interact directly with this functionality without providing valid credentials or session tokens. This represents a fundamental failure in the principle of least privilege and proper access control implementation, where administrative or state-changing operations are exposed to unauthenticated users.
From a technical perspective, the vulnerability permits an attacker to submit crafted token-revocation lists remotely. Token revocation is typically used to invalidate previously issued authentication credentials, thereby preventing their further use by legitimate users or services. By exploiting this flaw, an adversary can manipulate these lists arbitrarily. This capability allows the attacker to revoke tokens belonging to other valid users or system components without authorization. The ability to inject maliciously crafted data into this endpoint suggests that input validation and integrity checks are also likely insufficient, although the primary vector is the lack of authentication gating.
The operational impact of this vulnerability is severe, primarily manifesting as data tampering and denial of service conditions. Data tampering occurs because an attacker can alter the state of token validity, effectively changing who has access to resources or services managed by NVIDIA AIStore. This undermines the integrity of the system's security model. Furthermore, a denial of service condition arises when legitimate users are unable to perform their tasks because their valid tokens have been maliciously revoked by the attacker. In distributed computing environments like those supported by AIStore, such disruptions can halt critical machine learning workflows or data processing pipelines, leading to significant operational downtime and potential loss of productivity.
This vulnerability aligns with CWE-287, which describes Improper Authentication, as well as CWE-613, Insufficient Session Expiration, depending on the specific mechanics of how tokens are managed post-revocation. In terms of offensive security frameworks, this behavior maps to MITRE ATT&CK technique T1098, specifically Account Manipulation or SSH Hijacking if applied to session management contexts, as it involves altering authentication states to disrupt service or gain unauthorized access. The exploitation path falls under remote code execution vectors that do not require user interaction but rely on network accessibility and lack of input sanitization combined with missing auth checks.
To mitigate this risk, immediate remediation should focus on enforcing strict authentication requirements for the token-revocation endpoint. Only authenticated users with appropriate administrative privileges should be permitted to submit revocation requests. Additionally, implementing robust input validation is essential to ensure that submitted token lists are well-formed and originate from trusted sources. Network-level controls such as firewalls or API gateways can also restrict access to this specific endpoint to known internal IP ranges if the service does not need to be exposed publicly. Regular security audits and penetration testing of authentication mechanisms across all endpoints in NVIDIA AIStore will help identify similar gaps before they are exploited in production environments.