Speclad

Name of the Vulnerable Software and Affected Versions: CVAT versions 2.2.0 through 2.39.0 Description: CVAT is an open source interactive video and image annotation tool for computer vision. The issue arises from the lack of validation during the import process of a project or task backup, allowing an attacker with a CVAT account and a `user` role to potentially access and steal data by creating projects or tasks using files belonging to other users, if they know the filenames of those uploads. This issue does not affect annotation or dataset TUS uploads. Recommendations: For CVAT versions 2.2.0 through 2.39.0, upgrade to CVAT 2.40.0 or a later version to receive a patch. As a temporary workaround, consider restricting access to the import process of project or task backups to minimize the risk of exploitation.

**Name of the Vulnerable Software and Affected Versions** Computer Vision Annotation Tool (CVAT) versions 2.4.0 through 2.37.x **Description** The issue allows an authenticated user to retrieve sensitive information, including IDs and names of tasks, projects, labels, and IDs of jobs and quality reports, from a CVAT instance. This can lead to a denial of service if the instance contains a large number of resources, as retrieving this information may consume system resources and deny access to legitimate users. **Recommendations** For versions 2.4.0 through 2.37.x, update to version 2.38.0 to resolve the issue.

**Name of the Vulnerable Software and Affected Versions** Computer Vision Annotation Tool (CVAT) versions prior to 2.26.0 **Description** The issue allows an attacker with an account on an affected CVAT instance to run arbitrary code in the context of the Nuclio function container. This affects CVAT deployments that run serverless functions of type tracker from the CVAT Git repository, such as TransT and SiamMask. Deployments with custom functions of type tracker may also be affected if they use an unsafe serialization library like `pickle` or `jsonpickle`. **Recommendations** Upgrade to CVAT 2.26.0 or later. If you are unable to upgrade, shut down any instances of the TransT or SiamMask functions you're running.

**Name of the Vulnerable Software and Affected Versions** Computer Vision Annotation Tool (CVAT) versions prior to 2.19.0 **Description** The Computer Vision Annotation Tool (CVAT) is an interactive video and image annotation tool for computer vision. A malicious CVAT user with permissions to create or edit a task can trick another logged-in user into visiting a maliciously-constructed URL, allowing the attacker to initiate API calls on the victim's behalf and gain temporary access to the victim's data. **Recommendations** Upgrade to CVAT 2.19.0 or a later version to fix this issue. As a temporary workaround, consider restricting access to sensitive data and API endpoints to minimize the risk of exploitation. Avoid using maliciously-constructed URLs and be cautious when visiting links from untrusted sources.

**Name of the Vulnerable Software and Affected Versions** Computer Vision Annotation Tool (CVAT) versions prior to 2.18.0 **Description** The Computer Vision Annotation Tool (CVAT) is an interactive video and image annotation tool for computer vision. An attacker with a CVAT account can access webhook delivery information for any webhook registered on the CVAT instance, including that of other users. For each delivery, this contains information about the event that caused the delivery, typically including full details about the object on which an action was performed, and the user who performed the action. In addition, the attacker can redeliver any past delivery of any webhook, and trigger a ping event for any webhook. **Recommendations** Upgrade to CVAT 2.18.0 or any later version.

**Name of the Vulnerable Software and Affected Versions** Computer Vision Annotation Tool (CVAT) versions 2.1.0 through 2.14.3 **Description** The issue allows an attacker with a CVAT account to exploit a feature by specifying custom endpoint URLs for cloud storages based on Amazon S3 and Azure Blob Storage, potentially probing the network for HTTP(S) servers and creating a cloud storage linked to an internal server. This could enable the attacker to list files, extract files, and/or overwrite files on the server, depending on the internal server's configuration. **Recommendations** For Computer Vision Annotation Tool (CVAT) versions 2.1.0 through 2.14.3, upgrade to CVAT 2.14.3 to receive a patch, which applies existing SSRF mitigation measures to requests to cloud providers and prohibits access to intranet IP addresses by default. As a temporary workaround, consider using network security solutions such as virtual networks or firewalls to prohibit network access from the CVAT backend to unrelated servers on your internal network and/or require authentication for access to internal servers.

**Name of the Vulnerable Software and Affected Versions** Computer Vision Annotation Tool (CVAT) versions 2.2.0 through 2.14.3 **Description** The issue allows an attacker to initiate a dataset export or a backup from a project, task, or job that the victim user has permission to export into a cloud storage that the victim user has access to, by tricking a logged-in CVAT user into visiting a malicious URL. The attacker can choose the name of the resulting file, which implies they can overwrite arbitrary files in any cloud storage the victim can access. If the attacker has read access to the cloud storage used in the attack, they can obtain media files, annotations, settings, and other information from any projects, tasks, or jobs that the victim has permission to export. **Recommendations** For versions prior to 2.14.3, update to version 2.14.3 to resolve the issue. As a temporary workaround, consider restricting access to cloud storage for CVAT users to minimize the risk of exploitation. Avoid using CVAT to export datasets or backups to cloud storage until the issue is resolved.