CVE-2025-23045

Computer Vision Annotation Tool (CVAT) is an interactive video and image annotation tool for computer vision. An attacker with an account on an affected CVAT instance is able to run arbitrary code in the context of the Nuclio function container. This vulnerability affects CVAT deployments that run any of the serverless functions of type tracker from the CVAT Git repository, namely TransT and SiamMask. Deployments with custom functions of type tracker may also be affected, depending on how they handle state serialization. If a function uses an unsafe serialization library such as pickle or jsonpickle, its likely to be vulnerable. Upgrade to CVAT 2.26.0 or later. If you are unable to upgrade, shut down any instances of the TransT or SiamMask functions youre running.

Weakness

The product deserializes untrusted data without sufficiently ensuring that the resulting data will be valid.

Affected Software

Potential Mitigations

• Make fields transient to protect them from deserialization.
• An attempt to serialize and then deserialize a class containing transient fields will result in NULLs where the transient data should be. This is an excellent way to prevent time, environment-based, or sensitive variables from being carried over and used improperly.

Related Attack Patterns

• https://cwe.mitre.org/data/definitions/586.html

References

• https://github.com/cvat-ai/cvat/commit/563e1dfde64b15fa042b23f9d09cd854b35f0366
• https://github.com/cvat-ai/cvat/security/advisories/GHSA-wq36-mxf8-hv62