Veille CVE

‹ Retour à la liste

CVE-2026-5241

9.6 Critique

Publiée le 03/06/2026 à 16:16 · Modifiée le 28/08/2026 à 18:18 · Modifiée · Source : security@huntr.dev

Description

A vulnerability in the LightGlue model loading path of huggingface/transformers version 5.2.0 allows an attacker-controlled model repository to execute arbitrary code during model initialization. The issue arises because the `trust_remote_code` parameter, intended to prevent remote code execution, is overridden by untrusted serialized configuration data in a nested code path. Specifically, when loading a LightGlue model using `AutoModel.from_pretrained()` with `trust_remote_code=False`, the `LightGlueConfig` reads the `trust_remote_code` value from the untrusted `config.json` file and propagates it into nested `AutoConfig.from_pretrained()` calls. This results in the execution of attacker-provided Python modules, even when the victim explicitly disables remote code execution. The vulnerability poses a high risk for environments such as API inference servers, research notebooks, CI/CD pipelines, and model evaluation workers, potentially leading to credential theft, lateral movement, or persistence/backdoor deployment.

Traduire en français

Produits concernés

Scores CVSS

VersionScoreSourceVecteur
3.1 9.6 Critique nvd@nist.gov (NVD) CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H
3.1 7.7 Élevée 0b0ca135-0b70-47e7-9f44-1890c2a1c46c CVSS:3.1/AV:N/AC:H/PR:L/UI:R/S:C/C:H/I:H/A:N
3.0 8.0 Élevée security@huntr.dev CVSS:3.0/AV:N/AC:H/PR:N/UI:R/S:C/C:H/I:H/A:N

Faiblesses (CWE)

Références

Ailleurs