CVE-2025-68430 is a path traversal vulnerability (CWE-24) identified in the open-source CVAT (Computer Vision Annotation Tool) software, versions 2.8.1 through 2.52.0. CVAT is widely used for interactive video and image annotation in computer vision projects. The vulnerability allows an attacker who has an authenticated account on a CVAT instance to manipulate file path inputs to traverse directories on the server filesystem. This traversal enables the attacker to retrieve directory listings—names of files and subdirectories—within any directory accessible by the CVAT server process. Notably, the vulnerability does not permit reading file contents, limiting the exposure to metadata disclosure. The flaw arises from insufficient validation or sanitization of user-supplied path parameters, allowing '../' sequences to escape intended directory boundaries. The vulnerability is remotely exploitable without user interaction beyond authentication and requires low attack complexity. The CVSS 4.0 vector (AV:N/AC:L/PR:L/UI:N/VC:L/VI:N/VA:N/SC:N/SI:N/SA:N) indicates network attack vector, low complexity, privileges required (authenticated user), no user interaction, and limited confidentiality impact. The vulnerability was publicly disclosed on December 19, 2025, and patched in CVAT version 2.53.0. No known exploits are currently observed in the wild. Given CVAT's role in AI and computer vision workflows, this vulnerability could facilitate reconnaissance activities by exposing directory structures, potentially aiding further attacks or data leakage if combined with other vulnerabilities or misconfigurations.

For European organizations, the impact primarily involves information disclosure that could assist attackers in mapping the server filesystem, identifying sensitive directories, or locating configuration files and scripts. While direct file content access is not possible, knowledge of directory contents can facilitate targeted attacks such as privilege escalation, exploitation of other vulnerabilities, or social engineering. Organizations in sectors heavily invested in AI, autonomous systems, or research institutions using CVAT for annotation tasks are at higher risk. Exposure of directory structures could also reveal operational details or intellectual property indirectly. Since exploitation requires authenticated access, insider threats or compromised user accounts pose a significant risk. The vulnerability does not directly impact system availability or integrity but can be a stepping stone in multi-stage attacks. The medium CVSS score reflects moderate risk, but the absence of known exploits suggests limited immediate threat. However, delayed patching could increase exposure, especially in environments with weak account management or where CVAT instances are internet-facing.

The primary mitigation is to upgrade all CVAT instances to version 2.53.0 or later, where the vulnerability is patched. Since no workarounds exist, organizations should prioritize patch management. Additionally, implement strict access controls and monitoring on CVAT user accounts to prevent unauthorized access. Enforce strong authentication mechanisms, such as multi-factor authentication (MFA), to reduce the risk of account compromise. Network segmentation should be employed to limit CVAT server exposure to trusted internal networks only. Logging and alerting on unusual directory traversal attempts or abnormal user behavior can help detect exploitation attempts. Regularly audit CVAT configurations and user permissions to minimize privilege levels. If immediate upgrade is not feasible, consider restricting CVAT access via VPN or IP whitelisting to trusted users. Finally, educate users about the risks of credential compromise and enforce password hygiene.