Threats Tagged 'cve-2026-18618'

A path traversal and arbitrary file overwrite vulnerability exists in Argo Workflows used within Red Hat OpenShift AI. This flaw allows an attacker with workflow submission privileges to submit a malicious archive containing symbolic links that bypass extraction directory restrictions, leading to overwriting files inside the workflow pod. Successful exploitation can result in arbitrary command execution during pod startup. The impact is limited to the compromised pod and does not extend to host-level compromise. No official fix or mitigation currently meets Red Hat's standards for deployment or stability.

Red Hat OpenShift AI version 3.4.3 addresses multiple security issues identified by CVE-2026-13717 and others. The advisory announces updated images for Red Hat OpenShift AI to mitigate these vulnerabilities. The update is critical and users are advised to upgrade their clusters following Red Hat's official documentation. No specific technical details or exploit information are provided in the advisory.

A vulnerability in ml-metadata's statically-linked gRPC stack allows an in-cluster attacker with network access to the MLMD pod to cause a denial of service by sending specially crafted HTTP/2 requests. This flaw exploits outdated components vulnerable to known HTTP/2 DoS issues, potentially crashing the MLMD pod and disrupting pipeline runs in the affected namespace.

A flaw was found in ml-metadata. The statically-linked gRPC stack in ml-metadata is outdated, making it vulnerable to known HTTP/2 denial of service (DoS) issues. An in-cluster attacker, with network access to the MLMD pod, could exploit these vulnerabilities by sending specially crafted HTTP/2 requests. This could lead to a denial of service by crashing the MLMD pod, disrupting all pipeline runs in the affected namespace.

A Regular Expression Denial of Service (ReDoS) vulnerability (CVE-2026-15154) was found in guardrails-detectors, a component of Red Hat OpenShift AI. This flaw allows a remote attacker to send specially crafted regular expressions to the public detection API, causing catastrophic backtracking. The result is a worker process consuming 100% CPU indefinitely, leading to denial of service for the guardrails-mediated large language model (LLM) pipeline. The vulnerability affects specific versions of Red Hat OpenShift AI and has a medium severity rating with a CVSS score of 6.5. Updated images and a fixed version 2.25.10 are available from Red Hat to address this issue.