CVE-2024-11392 is a deserialization of untrusted data vulnerability (CWE-502) found in the Hugging Face Transformers library, specifically impacting the MobileViTV2 model. The vulnerability stems from inadequate validation of user-supplied configuration files, which are deserialized without proper security checks. This flaw allows remote attackers to execute arbitrary code by crafting malicious configuration data that, when processed by the vulnerable library, triggers unsafe deserialization routines. Exploitation requires user interaction, such as the victim opening a malicious file or visiting a malicious webpage that supplies the crafted data. The vulnerability affects the confidentiality, integrity, and availability of the system by enabling attackers to run arbitrary commands in the context of the current user. The CVSS v3.0 base score is 7.5 (high), reflecting network attack vector, high impact on confidentiality, integrity, and availability, but requiring user interaction and high attack complexity. No patches or exploits in the wild are currently reported, but the risk remains significant due to the widespread use of Hugging Face Transformers in AI/ML applications. The vulnerability was assigned by ZDI (ZDI-CAN-24322) and published on November 22, 2024.

The vulnerability allows remote attackers to execute arbitrary code on affected systems, potentially leading to full compromise of the host environment under the privileges of the user running the Hugging Face Transformers library. This can result in unauthorized data access, data manipulation, or disruption of AI/ML services. Organizations relying on Hugging Face Transformers for AI workloads, especially those processing untrusted input or exposed to external users, face risks of data breaches, service outages, and lateral movement within networks. The requirement for user interaction limits automated exploitation but does not eliminate risk, especially in environments where users may open untrusted files or visit malicious sites. The impact extends to confidentiality, integrity, and availability, making this a critical concern for AI-driven applications and services.

1. Immediately update to a patched version of Hugging Face Transformers once available, as no official patch is currently listed. 2. Until a patch is released, restrict the processing of untrusted configuration files or inputs by implementing strict input validation and sanitization. 3. Employ application-level sandboxing or containerization to limit the privileges and impact of potential code execution. 4. Educate users to avoid opening untrusted files or visiting suspicious websites that could trigger exploitation. 5. Monitor network and application logs for unusual activity indicative of exploitation attempts. 6. Use endpoint protection solutions capable of detecting anomalous behaviors related to deserialization attacks. 7. Consider disabling or restricting features that automatically load or deserialize configuration files from untrusted sources. 8. Implement network-level controls to limit exposure of systems running Hugging Face Transformers to untrusted networks.