Data Science Pipelines - CraftedSignal Threat Feedhttps://feed.craftedsignal.io/vendors/data-science-pipelines/Trending threats, MITRE ATT&CK coverage, and detection metadata. Fed continuously.Hugoenhello@craftedsignal.iohello@craftedsignal.ioMon, 10 Aug 2026 21:39:03 +0000CVE-2026-18621 Privilege Escalation in Data Science Pipelineshttps://feed.craftedsignal.io/briefs/2026-08-dsp-cve-2026-18621/Mon, 10 Aug 2026 21:39:03 +0000hello@craftedsignal.iohttps://feed.craftedsignal.io/briefs/2026-08-dsp-cve-2026-18621/An attacker with namespace editor privileges can bypass security hardening via the V1 API to execute malicious Argo Workflows, resulting in node-root access.CVE-2026-18621 is a critical security vulnerability discovered in Data Science Pipelines (DSP). The flaw exists within the V1 API path, which fails to correctly enforce security hardening for submitted workflows. An authenticated user with existing namespace editor privileges can submit a specially crafted Argo Workflow that forces the DSP API server to act as a confused deputy. By manipulating the workflow submission, the attacker triggers the creation of pods that inherit unauthorized elevated privileges. If successfully exploited, the attacker gains the ability to execute arbitrary code with node-root access, effectively compromising the integrity of the underlying Kubernetes node hosting the pipeline components. This vulnerability represents a significant risk for multi-tenant environments where namespace isolation is a primary security control.

Impact

Successful exploitation of this vulnerability results in a total compromise of the affected Kubernetes node. An attacker can escape the restricted namespace environment to gain node-root access, allowing for data exfiltration, lateral movement within the cluster, and persistent access to the infrastructure. Organizations utilizing DSP in shared or multi-tenant environments are at the highest risk, as the vulnerability circumvents standard RBAC restrictions assigned to namespace editors.

Recommendation

  • Audit all active Argo Workflow definitions and pipeline configurations to ensure they originate from trusted sources.
  • Apply the vendor-provided patch to Data Science Pipelines immediately to restrict V1 API workflow submission parameters.
  • Implement strict Pod Security Admissions (PSA) or policy enforcement controllers (e.g., OPA Gatekeeper or Kyverno) to prevent the creation of privileged pods within the namespace.
  • Monitor Kubernetes API server logs for unauthorized or unexpected workflow submissions targeting the V1 API path.
]]>highadvisory