Vulnerability-Lookup

GHSA-XRQW-3RRV-VX5W

Vulnerability from github – Published: 2026-08-02 18:30 – Updated: 2026-08-02 18:30

Details

A vulnerability in huggingface/transformers versions <=5.8.0.dev0 allows an attacker to perform arbitrary file writes via path traversal. The issue resides in the save_pretrained() methods of PreTrainedTokenizerBase and ProcessorMixin, where keys from the chat_template dictionary are used directly as filenames without proper validation. An attacker can exploit this by publishing a malicious Hugging Face Hub repository with a crafted tokenizer_config.json file. When a victim downloads and saves the tokenizer or processor, the attacker-controlled keys can escape the intended save directory, enabling arbitrary file writes with attacker-controlled content. This vulnerability affects multiple processors inheriting from ProcessorMixin, including Idefics, Florence, Gemma, Phi, and Qwen-VL.

Severity

7.1 (High)


                  
                    CVSS:3.0/AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:H/A:L

Show details on source website

JSON

To clipboard

{
  "affected": [],
  "aliases": [
    "CVE-2026-9856"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-22"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2026-08-02T16:16:25Z",
    "severity": "HIGH"
  },
  "details": "A vulnerability in huggingface/transformers versions \u003c=5.8.0.dev0 allows an attacker to perform arbitrary file writes via path traversal. The issue resides in the `save_pretrained()` methods of `PreTrainedTokenizerBase` and `ProcessorMixin`, where keys from the `chat_template` dictionary are used directly as filenames without proper validation. An attacker can exploit this by publishing a malicious Hugging Face Hub repository with a crafted `tokenizer_config.json` file. When a victim downloads and saves the tokenizer or processor, the attacker-controlled keys can escape the intended save directory, enabling arbitrary file writes with attacker-controlled content. This vulnerability affects multiple processors inheriting from `ProcessorMixin`, including Idefics, Florence, Gemma, Phi, and Qwen-VL.",
  "id": "GHSA-xrqw-3rrv-vx5w",
  "modified": "2026-08-02T18:30:20Z",
  "published": "2026-08-02T18:30:20Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-9856"
    },
    {
      "type": "WEB",
      "url": "https://github.com/huggingface/transformers/commit/eaaaf8494dd5386634ae37d1d122212fdc315be5"
    },
    {
      "type": "WEB",
      "url": "https://huntr.com/bounties/362824d5-fe18-40e8-a6cf-62277f97a170"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.0/AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:H/A:L",
      "type": "CVSS_V3"
    }
  ]
}

CVE-2026-9856 (GCVE-0-2026-9856)

Vulnerability from cvelistv5 – Published: 2026-08-02 15:10 – Updated: 2026-08-03 14:57

Title

Path Traversal in huggingface/transformers

Summary

A vulnerability in huggingface/transformers versions <=5.8.0.dev0 allows an attacker to perform arbitrary file writes via path traversal. The issue resides in the `save_pretrained()` methods of `PreTrainedTokenizerBase` and `ProcessorMixin`, where keys from the `chat_template` dictionary are used directly as filenames without proper validation. An attacker can exploit this by publishing a malicious Hugging Face Hub repository with a crafted `tokenizer_config.json` file. When a victim downloads and saves the tokenizer or processor, the attacker-controlled keys can escape the intended save directory, enabling arbitrary file writes with attacker-controlled content. This vulnerability affects multiple processors inheriting from `ProcessorMixin`, including Idefics, Florence, Gemma, Phi, and Qwen-VL.

Severity

7.1 (High)


                        
                          CVSS:3.0/AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:H/A:L

SSVC

Exploitation: none Automatable: no Technical Impact: partial

CISA Coordinator (v2.0.3)

CWE

CWE-22 - Improper Limitation of a Pathname to a Restricted Directory ('Path Traversal')

Assigner

@huntr_ai

References

2 references

URL	Tags
https://huntr.com/bounties/362824d5-fe18-40e8-a6c…
https://github.com/huggingface/transformers/commi…

Impacted products

1 product

Vendor	Product	Version
huggingface	huggingface/transformers	Affected: unspecified , < 5.10.0 (custom)

Show details on NVD website

JSON

To clipboard

{
  "containers": {
    "adp": [
      {
        "metrics": [
          {
            "other": {
              "content": {
                "id": "CVE-2026-9856",
                "options": [
                  {
                    "Exploitation": "none"
                  },
                  {
                    "Automatable": "no"
                  },
                  {
                    "Technical Impact": "partial"
                  }
                ],
                "role": "CISA Coordinator",
                "timestamp": "2026-08-03T14:57:31.978427Z",
                "version": "2.0.3"
              },
              "type": "ssvc"
            }
          }
        ],
        "providerMetadata": {
          "dateUpdated": "2026-08-03T14:57:43.370Z",
          "orgId": "134c704f-9b21-4f2e-91b3-4a467353bcc0",
          "shortName": "CISA-ADP"
        },
        "title": "CISA ADP Vulnrichment"
      }
    ],
    "cna": {
      "affected": [
        {
          "product": "huggingface/transformers",
          "vendor": "huggingface",
          "versions": [
            {
              "lessThan": "5.10.0",
              "status": "affected",
              "version": "unspecified",
              "versionType": "custom"
            }
          ]
        }
      ],
      "descriptions": [
        {
          "lang": "en",
          "value": "A vulnerability in huggingface/transformers versions \u003c=5.8.0.dev0 allows an attacker to perform arbitrary file writes via path traversal. The issue resides in the `save_pretrained()` methods of `PreTrainedTokenizerBase` and `ProcessorMixin`, where keys from the `chat_template` dictionary are used directly as filenames without proper validation. An attacker can exploit this by publishing a malicious Hugging Face Hub repository with a crafted `tokenizer_config.json` file. When a victim downloads and saves the tokenizer or processor, the attacker-controlled keys can escape the intended save directory, enabling arbitrary file writes with attacker-controlled content. This vulnerability affects multiple processors inheriting from `ProcessorMixin`, including Idefics, Florence, Gemma, Phi, and Qwen-VL."
        }
      ],
      "metrics": [
        {
          "cvssV3_0": {
            "attackComplexity": "LOW",
            "attackVector": "NETWORK",
            "availabilityImpact": "LOW",
            "baseScore": 7.1,
            "baseSeverity": "HIGH",
            "confidentialityImpact": "NONE",
            "integrityImpact": "HIGH",
            "privilegesRequired": "NONE",
            "scope": "UNCHANGED",
            "userInteraction": "REQUIRED",
            "vectorString": "CVSS:3.0/AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:H/A:L",
            "version": "3.0"
          }
        }
      ],
      "problemTypes": [
        {
          "descriptions": [
            {
              "cweId": "CWE-22",
              "description": "CWE-22 Improper Limitation of a Pathname to a Restricted Directory (\u0027Path Traversal\u0027)",
              "lang": "en",
              "type": "CWE"
            }
          ]
        }
      ],
      "providerMetadata": {
        "dateUpdated": "2026-08-02T15:10:53.809Z",
        "orgId": "c09c270a-b464-47c1-9133-acb35b22c19a",
        "shortName": "@huntr_ai"
      },
      "references": [
        {
          "url": "https://huntr.com/bounties/362824d5-fe18-40e8-a6cf-62277f97a170"
        },
        {
          "url": "https://github.com/huggingface/transformers/commit/eaaaf8494dd5386634ae37d1d122212fdc315be5"
        }
      ],
      "source": {
        "advisory": "362824d5-fe18-40e8-a6cf-62277f97a170",
        "discovery": "EXTERNAL"
      },
      "title": "Path Traversal in huggingface/transformers"
    }
  },
  "cveMetadata": {
    "assignerOrgId": "c09c270a-b464-47c1-9133-acb35b22c19a",
    "assignerShortName": "@huntr_ai",
    "cveId": "CVE-2026-9856",
    "datePublished": "2026-08-02T15:10:53.809Z",
    "dateReserved": "2026-05-28T15:41:24.076Z",
    "dateUpdated": "2026-08-03T14:57:43.370Z",
    "state": "PUBLISHED"
  },
  "dataType": "CVE_RECORD",
  "dataVersion": "5.2"
}

Sightings

Author	Source	Type	Date	Other

Nomenclature

Seen: The vulnerability was mentioned, discussed, or observed by the user.
Confirmed: The vulnerability has been validated from an analyst's perspective.
Published Proof of Concept: A public proof of concept is available for this vulnerability.
Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
Not confirmed: The user expressed doubt about the validity of the vulnerability.
Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.

Detection rules are retrieved from Rulezet.

The MITRE ATT&CK techniques below are AI-generated suggestions, inferred from the description of the vulnerability by the CIRCL/vulnerability-attack-technique-classification-roberta-base model, served locally by ML-Gateway. They have not been verified by an analyst and are provided for guidance only.

Action not permitted

GHSA-XRQW-3RRV-VX5W

CVE-2026-9856 (GCVE-0-2026-9856)

Tags

Sightings

Nomenclature