Executive brief
Doccano is an open-source annotation tool used by machine learning teams to label training data. A flaw in the bulk-delete endpoint allows attackers to bypass access controls and manipulate category labels through the API without proper authorization checks, potentially compromising data integrity in ML annotation projects.
Technical details
The vulnerability is an improper access control flaw in the bulk-delete endpoint (/v1/projects/1/category-types) that affects the LabelList function. The attack is network-reachable and does not require authentication, allowing remote attackers to perform unauthorized manipulation of category type data. An attacker can exploit this to delete, modify, or access labels and annotations they should not have permission to manage, compromising the integrity of annotation datasets. The flaw affects Doccano up to version 1.8.5; the vendor has not responded to disclosure efforts.
Affected products
- Doccano Doccano up to 1.8.5
Timeline
- 2026-08-31: disclosed
- other: Exploit released to public