CVE-2026-100975 in OpenShellinfo

Summary

by MITRE • 11/17/2026

NVIDIA OpenShell for Linux contains a vulnerability in the Kubernetes sandbox SSH authentication configuration where a local attacker with low privileges could obtain sensitive authentication information by accessing an SSH handshake secret stored in a plaintext environment variable and reuse the shared secret to gain unauthorized access to other sandboxes. A successful exploit of this vulnerability might lead to information disclosure, data tampering, and denial of service.

Be aware that VulDB is the high quality source for vulnerability data.

Analysis

by VulDB Data Team • 11/17/2026

The NVIDIA OpenShell for Linux platform incorporates a Kubernetes-based sandboxing architecture designed to provide isolated execution environments for developers and researchers. Within this infrastructure, secure shell access is facilitated through an SSH authentication mechanism that relies on shared secrets established during the session handshake process. A critical security flaw has been identified in how these cryptographic credentials are managed within the containerized environment. Specifically, the system stores sensitive authentication information, such as SSH handshake secrets or private keys required for establishing connections to sandbox instances, directly inside plaintext environment variables rather than utilizing more secure storage mechanisms like Kubernetes Secrets with encryption at rest or dedicated vault solutions. This architectural decision exposes critical security material to any process running within the same container context that has read access to the environment configuration.

This vulnerability is classified as CWE-312, which pertains to Cleartext Storage of Sensitive Information. The root cause lies in the improper handling and storage of credentials during the initialization phase of the sandboxed session. By exposing these secrets via standard environment variables, the system violates fundamental principles of least privilege and secure credential management. In a typical Kubernetes deployment, while pods are isolated from each other at the network level, they share the underlying node's kernel and can often access shared memory or process environments if proper security contexts like read-only root filesystems or restricted service accounts are not strictly enforced. Consequently, any local user with low privileges inside one sandbox instance can inspect the environment variables of related processes or containers to extract these plaintext secrets.

From an offensive perspective, this flaw aligns closely with MITRE ATT&CK technique T1098.004, which describes SSH Authorized Key Manipulation, and more broadly with lateral movement techniques where attackers leverage stolen credentials to move between systems. An attacker who successfully exploits this vulnerability can extract the shared secret or private key associated with a specific sandbox instance. With these credentials in hand, the adversary is not limited to accessing only their own isolated environment. They can reuse the extracted SSH handshake secrets to authenticate into other sandboxes within the same cluster that share similar configuration patterns or trust relationships. This capability effectively breaks the isolation boundary intended by the Kubernetes namespace and pod security policies, allowing a low-privilege user to escalate their access across multiple virtualized environments.

The operational impact of this vulnerability is severe and multifaceted. The primary consequence is unauthorized information disclosure, as sensitive cryptographic material is exposed without encryption or proper access controls. Beyond mere data leakage, the ability to reuse these credentials enables significant data tampering capabilities. An attacker can infiltrate other sandboxes to modify source code, inject malicious payloads into build pipelines, or alter configuration files that affect downstream deployments. Furthermore, this lack of isolation facilitates denial of service attacks against critical infrastructure components managed within those compromised sandboxes. The attacker could disrupt ongoing development workflows, corrupt shared resources, or use the foothold in one sandbox as a pivot point to target more sensitive parts of the NVIDIA OpenShell ecosystem, potentially leading to broader system compromise if further privilege escalation vectors exist.

To mitigate this vulnerability, immediate remediation steps must focus on eliminating plaintext storage of credentials within environment variables. The most effective solution is to migrate all SSH keys and handshake secrets into Kubernetes Secrets objects with encryption at rest enabled via a provider such as AWS KMS or Azure Key Vault. Additionally, application-level changes should ensure that these secrets are injected securely only when necessary for the specific process requiring them, rather than being broadly available in the environment scope. Implementing strict Pod Security Standards (PSS) and restricting access to secret volumes using RBAC policies will further limit exposure. Regular auditing of container configurations against CIS Kubernetes Benchmarks is also recommended to detect similar misconfigurations before they can be exploited by malicious actors seeking to breach sandbox isolation boundaries.

Responsible

Nvidia

Reservation

09/27/2026

Disclosure

11/17/2026

Moderation

accepted

CPE

ready

EPSS

0.00000

KEV

no

Activities

very low

Sources

Might our Artificial Intelligence support you?

Check our Alexa App!