/

CVE-2026-50540

Severity:

CRITICAL RISK

(9.6)

Path Traversal (CWE-22) in Kata Containers VM Runtime

CVE-2026-50540

Received

Vulnerability type

Category:

Path Traversal

CWE:

CWE-20, CWE-22

Impact:

Host Code Execution and Virtualization Boundary Escape

Severity:

CRITICAL RISK

(9.6)

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

A critical path validation flaw in Kata Containers allows a pod user to influence which host-side TOML configuration file the runtime loads. If the user can place a file at a host-visible path, the configuration may select attacker-controlled virtualization binaries and execute code with host-root privileges.

Affected context

The issue occurs in Kata Containers deployments using kata-runtime versions before 4.0.0 when pod annotations can specify an io.katacontainers.config_path value. Risk exists where a pod user can create or influence files at paths visible to the host and the runtime loads those files without restricting their location or validating their contents.

Exploit preconditions

  • Attack surface: The kata-runtime pod annotation and host-side configuration

loading path

  • Required access: Ability to create or place a file at a host-visible path

and influence the relevant pod annotation

  • Trigger point: Runtime processing of the configured TOML file during pod

startup or configuration

  • Why exploitation is possible: The annotation accepts an unrestricted path,

and the loaded configuration can select executable hypervisor or virtio-fs

daemon binaries

Why this happens

The underlying flaw is insufficient validation of a user-influenced configuration path. The runtime treats the annotation as a trusted reference to a host file instead of constraining it to an approved directory and validating the resulting configuration. This combines CWE-20, Improper Input Validation, with CWE-22, Path Traversal, because an attacker-controlled path can escape intended configuration boundaries and affect privileged executable selection.

Potential impact

  • An attacker may execute arbitrary code as root on the host.
  • A compromised pod could undermine the VM or container isolation boundary.
  • Host files, workloads, and credentials could be accessed or modified.
  • An attacker could use the host compromise for lateral movement or persistence.
  • Cluster integrity and availability could be affected.

General mitigation guidance

Mitigation generally focuses on removing trust from pod-supplied configuration paths and protecting privileged executable selection. Secure-by-default patterns include allowing only predefined configuration identifiers or immutable, administrator-controlled paths; canonicalizing and validating paths against an approved directory; applying strict TOML schemas and executable allowlists; and rejecting unexpected annotations. Host and runtime components should operate with least privilege, while file creation and mount policies should reduce pod access to host-visible locations.

How teams detect issues like this

Detection can include static analysis and code review rules for annotations or APIs that directly construct host paths from pod-controlled input. Review signals include missing canonicalization, unrestricted file reads, absent allowlists for binaries, and configuration parsing performed before policy checks. Runtime monitoring can alert on unusual io.katacontainers.config_path values, host-visible file creation by pods, and unexpected hypervisor or virtio-fs daemon execution. Admission policies can restrict the annotation and enforce approved runtime configurations.

Sources

SourceLink
github.comgithub.com/kata-containers/kata-containers/commit/03cc670076099530f4e1e9cb22849afdafb20f65
github.comgithub.com/kata-containers/kata-containers/security/advisories/GHSA-mp2j-xm59-qfgw

Frequently asked

questions

It can be exploitable through a pod or cluster control surface when an attacker

can influence the relevant annotation and place a file at a host-visible path.

It is not necessarily reachable by an unauthenticated external network user.

The attacker generally needs the ability to create or control a pod and write

a file where the host-side runtime can access it, subject to deployment

permissions and storage configuration.

Do not treat pod-provided paths as trusted. Prefer administrator-controlled

configuration identifiers or immutable allowlisted paths, and validate both

the resolved file location and permitted executable selections.

Enforce annotation policies, restrict host-visible file access, use least

privilege, apply strict configuration schemas, and monitor privileged runtime

binary execution.

Find out what's already
exploitable in your codebase.