/

CVE-2026-65091

Severity:

HIGH RISK

(8.8)

OS Command Injection (CWE-78) in NVIDIA OpenShell Gateway

CVE-2026-65091

Analyzed

Vulnerability type

Category:

OS Command Injection

CWE:

CWE-78

Impact:

Code Execution, Data Tampering, and Information Disclosure

Severity:

HIGH RISK

(8.8)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

NVIDIA OpenShell contains an OS command injection vulnerability in which a malicious gateway may influence command execution on a connected system. Successful exploitation could allow code execution, data tampering, and disclosure of sensitive information.

Affected context

The vulnerability may arise in NVIDIA OpenShell deployments that accept or process gateway-controlled data in an operating-system command execution path. Risk depends on whether an attacker can operate, compromise, or otherwise influence a gateway that communicates with the affected OpenShell component.

Exploit preconditions

  • Attack surface: Gateway-to-OpenShell communication or processing
  • Required access: Ability to control, compromise, or send crafted data through a malicious gateway
  • Trigger point: Handling of gateway-provided data in an OS command execution context
  • Why exploitation is possible: Untrusted input may be incorporated into a command without sufficiently separating data from executable command syntax

Why this happens

The underlying issue is an OS command injection flaw, mapped to CWE-78. The affected behavior may pass gateway-influenced data into a system command interpreter without reliably treating that data as inert arguments. When command construction and input handling are not safely separated, attacker-controlled characters or values can alter the intended command behavior.

Potential impact

  • An attacker may execute unauthorized operating-system commands
  • Application or host data could be modified or deleted
  • Sensitive information may be disclosed
  • The affected environment could be used for follow-on access or lateral movement

General mitigation guidance

Mitigation generally focuses on removing shell interpretation from application workflows and reducing trust in gateway-supplied data.

Common defensive patterns include:

  • Prefer direct process APIs that pass fixed argument arrays instead of invoking a command shell
  • Use strict allowlists for commands, operations, and accepted argument values
  • Apply context-appropriate validation and canonicalization before processing gateway input
  • Keep gateway and service privileges limited to the minimum required operations
  • Isolate command-executing components and restrict their filesystem, network, and process permissions
  • Avoid embedding untrusted values in dynamically constructed command strings

How teams detect issues like this

Detection approaches may include:

  • Static analysis rules for shell invocation APIs, dynamic command construction, and string concatenation near process execution
  • Code review checks for gateway-controlled values reaching command runners without allowlist validation
  • Runtime monitoring of unexpected child processes, shell launches, command arguments, and outbound connections
  • Policy enforcement that restricts shell execution and requires approved command interfaces
  • Automated security review tools that flag untrusted data flows into operating-system command execution

Sources

SourceLink
github.comgithub.com/NVIDIA/product-security/tree/main/2026/5872
nvd.nist.govnvd.nist.gov/vuln/detail/CVE-2026-65091
cve.orgcve.org/CVERecord?id=CVE-2026-65091

Frequently asked

questions

Exploitability depends on the deployment and gateway exposure. A malicious or compromised gateway must be able to supply data that reaches the vulnerable command-processing path.

The key prerequisite is control or influence over a gateway connection; whether authentication is required depends on the OpenShell deployment's access controls and trust configuration.

Avoid shell-based command construction, use direct process invocation with fixed arguments, and apply strict allowlists to gateway-controlled values.

Treat gateway input as untrusted, enforce least privilege, review data flows into process execution, and monitor for unexpected command or child-process activity.

Find out what's already
exploitable in your codebase.