EUVD-2025-208968
GHSA-5pqx-p8mp-cxrf
CVSS Vector
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
Lifecycle Timeline
3Description
NVIDIA Triton Inference Server Sagemaker HTTP server contains a vulnerability where an attacker may cause an exception. A successful exploit of this vulnerability may lead to denial of service.
Analysis
NVIDIA Triton Inference Server's Sagemaker HTTP server contains a race condition vulnerability that allows unauthenticated remote attackers to trigger an exception, resulting in denial of service. The vulnerability affects NVIDIA Triton Inference Server deployments using the Sagemaker HTTP server component and can be exploited over the network without authentication or user interaction. There is no indication of active exploitation (not in CISA KEV), and EPSS data was not provided, but the CVSS score of 7.5 (High) reflects the ease of exploitation.
Technical Context
NVIDIA Triton Inference Server is an open-source inference serving software that enables deployment of AI models at scale. The vulnerable component is the Sagemaker HTTP server, which provides an HTTP interface compatible with AWS SageMaker for model inference requests. The root cause is CWE-362 (Concurrent Execution using Shared Resource with Improper Synchronization, also known as a Race Condition), indicating that multiple threads or processes access shared resources without proper locking mechanisms, leading to undefined behavior and exceptions. The affected product CPE is cpe:2.3:a:nvidia:triton_inference_server:*:*:*:*:*:*:*:* with the wildcard version indicating multiple versions are vulnerable pending specific version guidance from NVIDIA's advisory.
Affected Products
NVIDIA Triton Inference Server versions utilizing the Sagemaker HTTP server component are affected, as identified by CPE cpe:2.3:a:nvidia:triton_inference_server:*:*:*:*:*:*:*:*. The wildcard version field indicates multiple versions are vulnerable, with specific version ranges detailed in NVIDIA's security bulletin at https://nvidia.custhelp.com/app/answers/detail/a_id/5790. Organizations should consult the official NVIDIA advisory for precise version information and confirmation of their deployment's vulnerability status.
Remediation
Consult NVIDIA's security bulletin at https://nvidia.custhelp.com/app/answers/detail/a_id/5790 for specific patch versions and upgrade instructions to remediate this race condition in the Sagemaker HTTP server component. Organizations should prioritize upgrading internet-facing Triton Inference Server instances to the patched version as soon as possible. As interim mitigation until patching is complete, restrict network access to the Sagemaker HTTP server endpoint using firewall rules or network segmentation to allow only trusted IP ranges, implement rate limiting to reduce the window for race condition exploitation, and monitor server logs for unusual exception patterns or service crashes that may indicate exploitation attempts.
Priority Score
Share
External POC / Exploit Code
Leaving vuln.today