« Volver al listado

CVE-2026-4372

Estado: AnalizadaAlta (7.8)—

A critical remote code execution vulnerability exists in all versions of the HuggingFace transformers library prior to version 5.3.0. The vulnerability allows an attacker to craft a malicious `config.json` file containing the `_attn_implementation_internal` field set to an attacker-controlled HuggingFace Hub repository ID. When a victim loads this model using the standard `AutoModelForCausalLM.from_pretrained()` API, the library downloads and executes arbitrary Python code from the attacker's repository with the victim's full OS privileges.

Leer descripción completaMostrar menos

This issue arises due to unfiltered deserialization of configuration attributes, insufficient sanitization of internal fields, and unsandboxed execution of downloaded kernels. The vulnerability bypasses the `trust_remote_code` security mechanism, is invisible to the victim, and exploits the standard documented usage pattern, making it particularly severe. Users are advised to upgrade to version 5.3.0 or later to mitigate this issue.

CVSS

Probabilidad de explotación (EPSS)

EPSS (Exploit Prediction Scoring System, de FIRST) estima la probabilidad de que una vulnerabilidad sea explotada en 30 días. Complementa a CVSS (impacto) y a CISA KEV (explotación confirmada).

🎯 Técnicas ATT&CK

Cómo se explota esta vulnerabilidad y qué consigue el atacante, en el lenguaje de MITRE ATT&CK.

AV:L UI:R indica ejecución en cliente (T1203); desserialización maliciosa de config.json ejecuta código Python arbitrario (T1059) con privilegios del usuario (T1068).

Inferido por nuestro agente de análisis a partir de la descripción oficial, el vector CVSS y la CWE, y comprobado por un supervisor. Puede contener errores.

🛡️ Mitigaciones ATT&CK que cubren estas técnicas

Execution PreventionApplication Isolation and SandboxingExploit ProtectionUpdate SoftwareThreat Intelligence ProgramRestrict Web-Based Content

Tecnologías afectadas (1)

Huggingface — Transformers

CWE

Referencias

JSON original (NVD)

Mostrar
{
  "id": "CVE-2026-4372",
  "cveTags": [],
  "metrics": {
    "ssvcV203": [
      {
        "source": "134c704f-9b21-4f2e-91b3-4a467353bcc0",
        "ssvcData": {
          "id": "CVE-2026-4372",
          "role": "CISA Coordinator",
          "options": [
            {
              "exploitation": "none"
            },
            {
              "automatable": "no"
            },
            {
              "technicalImpact": "total"
            }
          ],
          "version": "2.0.3",
          "timestamp": "2026-05-26T15:08:29.632933Z"
        }
      }
    ],
    "cvssMetricV30": [
      {
        "type": "Secondary",
        "source": "security@huntr.dev",
        "cvssData": {
          "scope": "UNCHANGED",
          "version": "3.0",
          "baseScore": 7.8,
          "attackVector": "LOCAL",
          "baseSeverity": "HIGH",
          "vectorString": "CVSS:3.0/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H",
          "integrityImpact": "HIGH",
          "userInteraction": "REQUIRED",
          "attackComplexity": "LOW",
          "availabilityImpact": "HIGH",
          "privilegesRequired": "NONE",
          "confidentialityImpact": "HIGH"
        },
        "impactScore": 5.9,
        "exploitabilityScore": 1.8
      }
    ]
  },
  "affected": [
    {
      "source": "security@huntr.dev",
      "affectedData": [
        {
          "vendor": "huggingface",
          "product": "huggingface/transformers",
          "versions": [
            {
              "status": "affected",
              "version": "unspecified",
              "lessThan": "5.3.0",
              "versionType": "custom"
            }
          ]
        }
      ]
    }
  ],
  "published": "2026-05-24T14:16:16.917",
  "references": [
    {
      "url": "https://github.com/huggingface/transformers/commit/a7f8e7ff37d87d1a1a0c8cf607971c607741452f",
      "tags": [
        "Patch"
      ],
      "source": "security@huntr.dev"
    },
    {
      "url": "https://huntr.com/bounties/1f693a6e-6836-4b8b-a0bd-ca036fba8884",
      "tags": [
        "Exploit",
        "Third Party Advisory"
      ],
      "source": "security@huntr.dev"
    }
  ],
  "vulnStatus": "Analyzed",
  "weaknesses": [
    {
      "type": "Secondary",
      "source": "security@huntr.dev",
      "description": [
        {
          "lang": "en",
          "value": "CWE-1066"
        }
      ]
    },
    {
      "type": "Primary",
      "source": "nvd@nist.gov",
      "description": [
        {
          "lang": "en",
          "value": "CWE-502"
        }
      ]
    }
  ],
  "descriptions": [
    {
      "lang": "en",
      "value": "A critical remote code execution vulnerability exists in all versions of the HuggingFace transformers library prior to version 5.3.0. The vulnerability allows an attacker to craft a malicious `config.json` file containing the `_attn_implementation_internal` field set to an attacker-controlled HuggingFace Hub repository ID. When a victim loads this model using the standard `AutoModelForCausalLM.from_pretrained()` API, the library downloads and executes arbitrary Python code from the attacker's repository with the victim's full OS privileges. This issue arises due to unfiltered deserialization of configuration attributes, insufficient sanitization of internal fields, and unsandboxed execution of downloaded kernels. The vulnerability bypasses the `trust_remote_code` security mechanism, is invisible to the victim, and exploits the standard documented usage pattern, making it particularly severe. Users are advised to upgrade to version 5.3.0 or later to mitigate this issue."
    },
    {
      "lang": "es",
      "value": "Existe una vulnerabilidad crítica de ejecución remota de código en todas las versiones de la biblioteca HuggingFace transformers anteriores a la versión 5.3.0. La vulnerabilidad permite a un atacante crear un archivo 'config.json' malicioso que contiene el campo '_attn_implementation_internal' configurado con un ID de repositorio de HuggingFace Hub controlado por el atacante. Cuando una víctima carga este modelo utilizando la API estándar 'AutoModelForCausalLM.from_pretrained()', la biblioteca descarga y ejecuta código Python arbitrario del repositorio del atacante con los privilegios completos del sistema operativo de la víctima. Este problema surge debido a la deserialización no filtrada de atributos de configuración, la sanitización insuficiente de campos internos y la ejecución sin sandboxing de kernels descargados. La vulnerabilidad elude el mecanismo de seguridad 'trust_remote_code', es invisible para la víctima y explota el patrón de uso estándar documentado, lo que la hace particularmente grave. Se aconseja a los usuarios actualizar a la versión 5.3.0 o posterior para mitigar este problema."
    }
  ],
  "lastModified": "2026-07-23T17:10:00.123",
  "configurations": [
    {
      "nodes": [
        {
          "negate": false,
          "cpeMatch": [
            {
              "criteria": "cpe:2.3:a:huggingface:transformers:*:*:*:*:*:*:*:*",
              "vulnerable": true,
              "matchCriteriaId": "DCF55C52-9279-4B31-A8E9-004C31DA635F",
              "versionEndExcluding": "5.3.0"
            }
          ],
          "operator": "OR"
        }
      ]
    }
  ],
  "sourceIdentifier": "security@huntr.dev"
}