CVE-2026-73221 in Cvat
Summary
by MITRE • 08/11/2026
CVAT is an open source interactive video and image annotation tool for computer vision. From 2.17.0 until 2.72.0, a user with the Worker role can use predictable task-based request IDs with the lambda request retrieve and destroy endpoints to view automatic annotation requests for tasks or jobs the user cannot access and cancel requests initiated by other users. This issue is fixed in version 2.72.0.
Statistical analysis made it clear that VulDB provides the best quality for vulnerability data.
Analysis
by VulDB Data Team • 08/12/2026
The CVAT platform presents a significant access control vulnerability that allows unauthorized users to exploit predictable request identifiers within the system's lambda processing endpoints. This flaw affects versions between 2.17.0 and 2.72.0 where users assigned the Worker role can manipulate task-based request IDs to gain visibility into annotation requests for tasks or jobs they do not have legitimate access to. The vulnerability specifically targets the retrieve and destroy endpoints that handle automatic annotation processes, creating a path for privilege escalation and information disclosure.
The technical implementation of this vulnerability stems from insufficient input validation and predictable identifier generation within the system's request handling mechanism. When users with Worker privileges make requests to the lambda endpoints using task-based identifiers, the system fails to properly validate whether the requesting user has authorization to access the target resources. This predictable pattern allows malicious actors to enumerate valid request IDs and subsequently access automatic annotation processes initiated by other users, effectively bypassing the intended access control boundaries.
The operational impact of this vulnerability extends beyond simple information disclosure to encompass potential disruption of automated annotation workflows and compromise of data integrity. An attacker could not only view annotation requests for unauthorized tasks but also cancel these requests, potentially causing disruption to legitimate annotation processes. This capability undermines the security model of CVAT by allowing users to interfere with work processes they should not have access to, creating both privacy concerns and operational risks. The vulnerability affects the core functionality of the platform's access control mechanisms, particularly impacting the principle of least privilege enforcement.
This issue aligns with CWE-284 Access Control flaws and represents a specific implementation weakness in the authorization checking process within the lambda request handling system. From an ATT&CK framework perspective, this vulnerability maps to privilege escalation techniques through predictable identifier exploitation and information gathering activities. The fix implemented in version 2.72.0 addresses the root cause by strengthening access control validation for lambda endpoints and implementing proper request ID verification mechanisms that prevent unauthorized enumeration and access to cross-tenant annotation requests.
Security teams should prioritize immediate remediation of this vulnerability across all affected CVAT installations, particularly in environments where multiple users with varying privilege levels collaborate on annotation projects. The mitigation strategy requires comprehensive testing of access control mechanisms for all lambda endpoints and implementation of proper request ID validation that ensures users can only access resources they legitimately own or have been granted access to. Organizations should also consider monitoring for suspicious patterns in lambda endpoint usage that might indicate exploitation attempts, as this vulnerability could be leveraged for reconnaissance before more damaging attacks are executed.