{
  "affected": [
    {
      "package": {
        "ecosystem": "Go",
        "name": "github.com/siderolabs/talos"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "1.12.7"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [],
  "database_specific": {
    "cwe_ids": [
      "CWE-669"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-05-07T02:38:02Z",
    "nvd_published_at": null,
    "severity": "HIGH"
  },
  "details": "### Summary\n\nA vulnerability in the Linux kernel\u0027s algif_aead subsystem (CVE-2026-31431, \"copy.fail\") allows an unprivileged container workload to corrupt arbitrary file page-cache pages via the AF_ALG crypto interface and splice(). On Talos Linux, this vulnerability can be chained into a complete node compromise: an attacker who can schedule a pod on a worker node can, without any elevated Kubernetes permissions, achieve arbitrary code execution as root on the host (by poisoning a binary inside a privileged pod, or poisoning a binary which runs with elevated privileges like a CNI binary), access host filesystem, including node secrets. \n\nThe exploit does not require kernel debugging, race conditions, or any prior privileges beyond the ability to create a pod.\n\n### Impact\n\nAn attacker with the ability to deploy a Kubernetes pod on an affected node can:\n\n1. Corrupt the page-cache of /usr/sbin/nft in the containerd snapshot layer shared between the attacker\u0027s pod and the kube-proxy DaemonSet. Because containerd reuses XFS page-cache pages across overlayfs mounts sharing the same lower layer, the corruption is immediately visible to all containers using that image layer \u2014 including privileged system DaemonSets.\n2. Execute arbitrary code inside kube-proxy \u2014 a privileged DaemonSet running on every node with all Linux capabilities (privileged: true) and host network access \u2014 the next time kube-proxy invokes nft as part of its nftables reconciliation loop (typically within seconds).\n3. At this point, an attacker achieved code execution inside a privileged pod, which allows to escape to the host.\n4. Same attack can be planted by infiltrating other binaries running as privileged, for example a CNI plugin.\n\n### Patches\n\nUpgrade to Talos v1.13.0 or Talos v1.12.7 which ships Linux kernel 6.18.25. The kernel fix for CVE-2026-31431 (algif_aead in-place optimization revert) was committed upstream in  Linux 6.18.22 and is included in all Talos releases from v1.13.0 and Talos 1.12.7 onwards.\n\n### Workarounds\n\nThere are multiple workarounds available based on the situation, but we really recommend to upgrade.\n\n#### Option 1 - Change kernel arguments\n\nAdd a kernel argument with `initcall_blacklist=algif_aead_init` by upgrading Talos to the same version.\n\n\u003e Note: this either requires setting `machine.kernel.extraKernelArgs` if using BIOS based boot or upgrading with a new image from factory/imager generated image by setting the extra kernel args. See [Boot Assets](https://docs.siderolabs.com/talos/v1.13/platform-specific-installations/boot-assets)\n\n#### Option 2 - Deploy all workload pods with a seccomp profile denying  creating `AF_ALG` socket creation\n\n`patch.yaml`\n\n```yaml\nmachine:\n  seccompProfiles:\n    - name: copy-fail-block.json\n      value:\n        defaultAction: SCMP_ACT_ALLOW\n        syscalls:\n          - names:\n              - socket\n            action: SCMP_ACT_ERRNO\n            args:\n              - index: 0\n                value: 38\n                op: SCMP_CMP_EQ\n```\n\nApply this patch to all machines in the cluster and set this for all the pod spec:\n\n```yaml\n...\nspec:\n      securityContext:\n        seccompProfile:\n          type: Localhost\n          localhostProfile: profiles/copy-fail-block.json\n```\n\n#### Option 3 - Block the syscall in runtime with a eBPF program\n\nSee [copy-fail-blocker](https://github.com/cozystack/copy-fail-blocker), this can be applied to a running system without a reboot, but it has to run before any other workloads are scheduled after a reboot. \n\n### References\n\n* https://copy.fail/\n* https://xint.io/blog/copy-fail-linux-distributions\n* https://github.com/theori-io/copy-fail-CVE-2026-31431\n* https://github.com/Percivalll/Copy-Fail-CVE-2026-31431-Kubernetes-PoC",
  "id": "GHSA-m38g-vww2-mvgx",
  "modified": "2026-06-08T23:43:27Z",
  "published": "2026-05-07T02:38:02Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/siderolabs/talos/security/advisories/GHSA-m38g-vww2-mvgx"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-31431"
    },
    {
      "type": "WEB",
      "url": "https://github.com/Percivalll/Copy-Fail-CVE-2026-31431-Kubernetes-PoC"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/siderolabs/talos"
    },
    {
      "type": "WEB",
      "url": "https://github.com/theori-io/copy-fail-CVE-2026-31431"
    },
    {
      "type": "WEB",
      "url": "https://xint.io/blog/copy-fail-linux-distributions"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:C/C:H/I:H/A:N",
      "type": "CVSS_V3"
    }
  ],
  "summary": "Talos Linux has a local privilege escalation from untrusted workloads"
}