What is the CVSS score of CVE-2026-47479?

CVE-2026-47479 has a CVSS 3.1 base score of 7.5 (High). CVSS vector: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H. EPSS exploitation probability: 0.3%.

Which versions of Triton Inference Server are affected by CVE-2026-47479?

NVIDIA Triton Inference Server for Linux is vulnerable to remote denial of service attacks that can completely exhaust server resources and disable AI/ML inference endpoints.

How to fix or mitigate CVE-2026-47479?

Within 24 hours, identify and inventory all NVIDIA Triton Inference Server instances, document version numbers, implement network-level rate limiting or Web Application Firewall rules to restrict incoming request rates, and activate real-time monitoring and alerting for resource utilization anomalies (CPU, memory, network I/O). Within 7 days, configure resource quotas at the per-user and per-session level including memory limits, connection limits, and request queue depth caps; deploy request authentication with API key-based rate throttling; implement connection timeouts to prevent resource exhaustion from idle connections; and conduct failover and recovery drills for each Triton deployment. Within 30 days, architect resilience improvements including load balancing across multiple Triton instances, implement progressive request degradation policies during resource stress, establish circuit breaker logic to reject requests when thresholds are exceeded, and create a patch-readiness process with automated deployment to apply vendor updates within 48 hours of release.

Triton Inference Server CVE-2026-47479

EUVD-2026-44431 HIGH

Uncontrolled Resource Consumption (CWE-400)

2026-07-14 nvidia

GHSA-6gfh-7rjr-vfmg

Denial Of Service Triton Inference Server Nvidia

7.5

CVSS 3.1 · Vendor: nvidia

Severity by source

Vendor (nvidia) PRIMARY

7.5 HIGH

AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H

vuln.today AI

7.5 HIGH

Network-reachable, unauthenticated, low-complexity DoS with availability-only impact; C/I set to N as the flaw is pure resource exhaustion.

3.1 AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H

4.0 AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N

Primary rating from Vendor (nvidia).

CVSS VectorVendor: nvidia

Attack Vector

Network

Attack Complexity

Low

Privileges Required

None

User Interaction

None

Scope

Unchanged

Confidentiality

None

Integrity

None

Availability

High

Lifecycle Timeline

Analysis Generated

Jul 14, 2026 - 21:02 vuln.today

CVE Published

Jul 14, 2026 - 19:46 cve.org

HIGH 7.5

DescriptionCVE.org

NVIDIA Triton Inference Server for Linux contains a vulnerability where an attacker can cause uncontrolled resource consumption. A successful exploit of this vulnerability might lead to denial of service.

Articles & Coverage 1

News 14 New Nvidia Vulnerabilities - 14 High Severity, No POC Available Jul 15, 2026

AnalysisAI

Denial of service in NVIDIA Triton Inference Server for Linux lets remote unauthenticated attackers exhaust server resources (CWE-400 uncontrolled resource consumption) and render the AI/ML inference endpoint unavailable, with no impact to confidentiality or integrity. The CVSS 3.1 base score is 7.5 (AV:N/AC:L/PR:N/UI:N/A:H), reflecting network-reachable, low-complexity, no-privilege exploitation. There is no public exploit identified at time of analysis and it is not listed in CISA KEV; no EPSS score was provided in the input.

Technical ContextAI

NVIDIA Triton Inference Server is an open-source model-serving platform that exposes HTTP/REST and gRPC endpoints for running deep-learning inference across frameworks (TensorRT, ONNX, PyTorch, TensorFlow, Python backends). The flaw is classified as CWE-400 (Uncontrolled Resource Consumption), the class of bug where an attacker-influenced input causes the server to allocate CPU, memory, threads, or other finite resources without an effective upper bound. In a model server this typically manifests when a client-supplied request (for example an inference request, tensor shape/size, or metadata payload) is processed before resource limits are enforced, allowing a single or repeated request to drive the process toward exhaustion. The affected CPE is cpe:2.3:a:nvidia:triton_inference_server with an unbounded version range, indicating the specific fixed version was not enumerated in the provided data.

RemediationAI

No vendor-released patch version was identified in the provided data, so consult the NVIDIA product security advisory for CVE-2026-47479 (referenced via https://nvd.nist.gov/vuln/detail/CVE-2026-47479) and upgrade Triton Inference Server to the fixed release NVIDIA specifies. As compensating controls until patched: do not expose the Triton HTTP (default 8000), gRPC (8001), and metrics (8002) ports to untrusted networks - place the server behind a reverse proxy or API gateway that enforces authentication and request-rate limiting, which trades a small latency/operational overhead for blocking anonymous request floods; restrict inbound access with firewall/security-group rules to known inference clients only; and apply resource ceilings at the container/orchestration layer (CPU/memory cgroup limits, request-size caps, and concurrency limits) so a single exhausting request degrades rather than crashes the host, at the cost of rejecting some legitimate large requests. These mitigations reduce reachability and blast radius but do not remove the underlying flaw, so patching remains the primary fix.